# Marketing trends 2027: The machines make the noise. Trust still makes the sale.

Seven stories from marketing history, from 1683 to the 1950s, explain the seven trends that will shape your 2027 plan. Each one ends with three actions you can start on Monday.

By Anže Frantar, Fractional CMO and Chief AI Officer, Ljubljana. Published 4 November 2026. 35-minute read.

Canonical page: <https://failagainfailbetter.com/marketing-2027/>

This is the Markdown version of the page, for AI tools and for anyone who prefers plain text. The page itself has the pictures and charts, the seven-question check, the calculator and a plan you can build action by action.

## Summary

Marketing 2027 is a free report by Anže Frantar on the seven marketing trends that will shape 2027 plans for B2B and B2C companies: AI search, buying groups, AI agents, proof, sameness, prices and green claims, and AI savings. Each chapter opens with a story from marketing history, from 1683 to the 1950s, answers one question with a short answer and the evidence from 2026, and ends with three actions for the plan, 21 in all. Every statistic links to its source, and most of the data comes from the US and Europe.

## Picture a planning meeting

The 2027 plan is on the screen. Someone near the window leans forward. "The new AI tools," he says, "save us 40% of our time."

The boss does not look up from her laptop when she answers. "Then we don't need the two new hires."

Nobody asks where the 40% comes from, because marketing has a long and comfortable habit of not asking. Its most famous line is usually credited to the Philadelphia retailer John Wanamaker: "Half the money I spend on advertising is wasted; the trouble is I don't know which half." The researchers at [Quote Investigator](https://quoteinvestigator.com/2022/04/11/advertising/) traced it back. Three dates stand out.

- **1890:** A partial version in the trade paper Printers' Ink
- **1898:** A partial version from Robert Ogden, who ran Wanamaker's New York store
- **1919:** The earliest credit to Wanamaker, in a preacher's speech

Wanamaker is still the most likely author. But for decades the best-known line about waste in our trade went from speech to speech, and nobody checked who had actually said it until much later.

This report follows seven changes that will shape next year's plan for companies that sell to other companies (B2B) and for those that sell to all of us (B2C). Each chapter opens with a story from marketing history, shows what 2026 taught us about the same problem and ends with three actions you can put in the plan on Monday.

Most of the data comes from the US and Europe, and I say so wherever a number is European. Every statistic links to its source, so you can do what nobody in that meeting room thought to do.

Ask.

**Hi, I'm Anže.** I help companies across Europe lead marketing and put AI to work. Usually that means joining their team as a [fractional CMO](https://failagainfailbetter.com/fractional-cmo/) or [Chief AI Officer](https://failagainfailbetter.com/chief-ai-officer/), often both.

I wrote this for the people who have to turn next year's plan into real work, and who would rather spend the evening at home than read another forty-page trend report full of charts. In every chapter you will find one box called "If I joined your team on Monday". That is what I would do first.

How this page was made: I used AI tools for research, drafting and fact-checking, and for the code behind the drawings and charts. Every statistic links to its source, so you can check it yourself. I take responsibility for the text.

## What are the marketing trends for 2027?

Seven trends will shape marketing plans in 2027, for companies that sell to other companies and for those that sell to consumers.

1. **AI search.** Fewer buyers will reach your website, so what the assistant says about you will matter more than where your page ranks. See [chapter 1](https://failagainfailbetter.com/marketing-2027/#c1).
2. **Buying groups.** Thirteen people inside the buying company and nine outside advisers sway a typical purchase, and the shortlist is often written before your salesperson calls. See [chapter 2](https://failagainfailbetter.com/marketing-2027/#c2).
3. **AI agents.** Agents already do the looking. For now, people still do the paying. See [chapter 3](https://failagainfailbetter.com/marketing-2027/#c3).
4. **Proof.** Buyers believe proof you cannot fake: what unpaid people say, and tests that cost you something. See [chapter 4](https://failagainfailbetter.com/marketing-2027/#c4).
5. **Sameness.** When everyone uses the same tools, the only thing that sets you apart is what is already yours. See [chapter 5](https://failagainfailbetter.com/marketing-2027/#c5).
6. **Prices and green claims.** Explain a price rise in the open and on the pack, and drop any green claim you cannot prove. See [chapter 6](https://failagainfailbetter.com/marketing-2027/#c6).
7. **AI savings.** AI saves far less time than the headlines say. My estimate: plan on getting back 3 to 7% of your hours. See [chapter 7](https://failagainfailbetter.com/marketing-2027/#c7).

### The 90-second version: seven numbers

Each number opens its chapter.

- **5 in 100** (chapter 1): ChatGPT sessions that send someone on to another website. For Google searches it is about 31 in 100. Source: Comscore US desktop data, October 2024 to July 2025, [arXiv paper, July 2026](https://arxiv.org/abs/2607.07652)
- **13 + 9** (chapter 2): People in a typical business purchase: 13 inside the buying company, and 9 outside advisers who sway the decision. Source: [Forrester, January 2026](https://www.forrester.com/press-newsroom/forrester-2026-the-state-of-business-buying/)
- **75%** (chapter 3): Business purchases in which AI agents will play a part in 2027, Forrester predicts. The final step usually stays with people. Source: [Forrester, 21 October 2026](https://www.forrester.com/press-newsroom/forresters-2027-b2b-go-to-market-predictions-ai-agents-will-influence-75-of-b2b-purchase-decisions/) (prediction)
- **41 vs 20** (chapter 4): Share of people who trust a brand most because of what unpaid people say, against the share who trust the brand's own words most. Source: [Edelman, June 2026](https://www.edelman.com/sites/g/files/aatuss191/files/2026-06/2026%20Trust%20and%20Brands%20-%20Top%2010%20Findings.pdf)
- **65%** (chapter 5): American consumers who say brands put out too much AI-made content. Source: [Gartner, September 2026](https://www.gartner.com/en/newsroom/press-releases/2026-09-22-gartner-marketing-survuey-finds-35-percent-of-consumers-rely-on-influencers-less-due-to-ai)
- **10 grams** (chapter 6): Removed from a chocolate bar whose pack looks almost the same. A German court called it misleading. The ruling is under appeal. Source: [Landgericht Bremen, May 2026](https://www.landgericht.bremen.de/sixcms/media.php/13/PM%2035-26%20Urteil%20im%20Verfahren%20Verbraucherzentrale%20gegen%20Mondelez.pdf) (under appeal)
- **2.2%** (chapter 7): Share of all US work hours that workers say generative AI saved them in the second quarter of 2026. The famous figure is 40%. Source: [Bick, Blandin and Deming, FRED Blog, August 2026](https://fred.stlouisfed.org/blog/2026/08/does-generative-ai-save-time-at-work/) (self-reported)

## Chapter 1: How will AI search change marketing in 2027?

**The machine that answers for you** (1900: The tyre maker that told France where to eat)

In 1900 France had fewer than 3,000 cars on its roads. The tyre maker Michelin printed a guide for their drivers anyway, and gave away 35,000 copies.

In 1920, the story goes, André Michelin found copies of the guide holding up a workbench. From then on the guide cost seven francs. The first stars came in 1926, [the guide's own history](https://guide.michelin.com/kr/en/article/features/history-michelin-guide) says. **A company that sold tyres became the judge of French kitchens for about a century.**

*The short answer:* **Fewer buyers will reach your website, so what the assistant says about you will matter more than where your page ranks.** In Comscore's US desktop data, about 5 in 100 ChatGPT sessions led on to another website, against 31 in 100 Google searches ([arXiv study](https://arxiv.org/abs/2607.07652), July 2026). ChatGPT reaches more than 1.2 billion weekly users, [OpenAI said](https://openai.com/index/gpt-6-for-everyone/) on 7 October 2026.

### The new guide answers the question itself

A guide that answers the question for you is bad news for anyone who runs a website. Researchers used Comscore browsing data to study what Americans actually did on their desktop computers after they asked a question.

**Chart:** Out of 100 ChatGPT sessions, about 5 send someone to another website. Out of 100 Google searches, about 31 do.

- Google search: 31 in 100
- ChatGPT session: 5 in 100

So a Google search leads on to a website about six times as often as a ChatGPT session. The ChatGPT share rose month by month from about 2.5% to a peak near 6.5% and stayed far below Google.

ChatGPT 5.2% and Google 31.1%: [arXiv 2607.07652](https://arxiv.org/abs/2607.07652), July 2026, from Comscore US desktop data, October 2024 to July 2025. A session and a search are not the same unit.

#### Where readers gather, ads follow

OpenAI began testing ads in the United States in February 2026 and brought them to 31 European countries in August. By 21 October more than 100,000 advertisers had run ads, [OpenAI said](https://openai.com/index/building-chatgpt-ad-stack/). It also said that in under 200 days ad sales had reached a rate of a billion dollars a year. Both figures are self-reported.

A week after the European launch, the European Commission [designated ChatGPT](https://ec.europa.eu/commission/presscorner/detail/en/ip_26_1772) a "very large online search engine". That designation puts ChatGPT under the EU's strictest platform rules from January 2027.

People use the new guide more than they trust it. Marketers are rushing in anyway. In [Kantar's survey](https://www.kantar.com/inspiration/advertising-media/media-reactions-2026) of 23,000 consumers in 33 markets and 800 senior marketers, published on 5 October, a third of people (33%) already research brands and products through AI assistants. Only 24% say they trust what those assistants recommend.

Even so, the marketers who plan to spend more on being seen there in 2027 outnumber those who plan to spend less by 75 points.

#### Does AI search send traffic? Almost none, so treat it as a brand channel

Tailwind Labs, which makes a toolkit for building websites, found this out early. Its toolkit was more popular than ever, but at the same time AI coding tools sent far fewer people to read its help pages. Visits to those pages fell 40% in two years and revenue fell by almost 80%, [its chief executive said](https://www.devclass.com/2026/01/08/tailwind-labs-lays-off-75-percent-of-its-engineers-thanks-to-brutal-impact-of-ai/) in January.

Two studies from October, each covering about 100 websites, found that AI assistants sent less than 1% of all visits. In one of them, ordinary search sent 39 times as much traffic.

[SparkToro](https://sparktoro.com/blog/is-ai-visibility-a-marketing-boondoggle/), citing Mike King, calls AI "more like a brand channel, a digital billboard if you will". People see your name there, but few come through to your site. So judge it by how often you are named, not by visits.

**Illustration on the page:** Same question, new answer every time.

In January 2026, SparkToro and Gumshoe collected 2,961 answers. Any two gave the same list of brands less than once in 100 tries. One check tells you nothing; a count over many runs does.

[SparkToro and Gumshoe](https://sparktoro.com/blog/new-research-ais-are-highly-inconsistent-when-recommending-brands-or-products-marketers-should-take-care-when-tracking-ai-visibility/), January 2026.

**Illustration on the page:** What a crawler sees on a typical pricing page.

With JavaScript off, the prices vanish, because code added them after the page loaded. If your prices vanish too, action 1 is a one-morning check.

In [Vercel's analysis](https://vercel.com/blog/the-rise-of-the-ai-crawler) from December 2024, the major AI crawlers, the programs that read pages for AI tools, did not run JavaScript. Google's crawler did. Behaviour changes, so test your own pages.

#### The board is already asking

Almost half of marketers (49%) already get questions from their executives or board about how the brand shows up in AI answers, [Forrester reported](https://www.forrester.com/blogs/predictions-2027-humans-reclaim-content-creation-machines-dominate-content-consumption/) on 30 October. Buyers still want a human at the end: in [Gartner's survey](https://www.gartner.com/en/newsroom/press-releases/2026-05-20-gartner-survey-finds-sixty-nine-percent-of-b-two-b-buyers-turn-to-sales-reps-to-validate-ai-generated-insights) of business buyers, 69% prefer to check what AI told them with a salesperson. The salesperson becomes the chatbot's fact-checker, and marketing now has to brief both of them.

> In 1926 a tyre maker started deciding where France ate. In 2027 the machine that answers decides who gets named.

**If I joined your team on Monday:** I would open your pricing page with JavaScript switched off, then sit down with the team, ask ChatGPT the ten questions your buyers ask most often and write down every brand that comes back. It takes one morning.

### Three actions for your 2027 plan, from chapter 1: The machine that answers for you

- **Action 1.** Switch JavaScript off and open your pricing and product pages. Rebuild whatever vanishes in plain HTML within 90 days. *What this means:* Plain HTML means the text sits in the page itself, so a program that reads pages for AI tools can see it without running code. Datasheets and comparison pages count too. *Tags:* costs time, B2B, B2C, web team.
- **Action 2.** Ask two assistants ten buyer questions, ten times each, every quarter. Count how often you and three rivals appear. *What this means:* Single answers change from run to run, so only a count means anything. Treat the result like a reputation score, not a traffic forecast. *Tags:* costs time, B2B, B2C, marketing.
- **Action 3.** Hand sales a one-page fact sheet: specs, references, how pricing works and what AI says about you now. *What this means:* Buyers bring what the assistant told them into the meeting. The salesperson needs the facts to confirm or correct it. *Tags:* costs time, B2B, sales.

## Chapter 2: How do business buyers choose suppliers in 2027?

**The list in the buyer's head** (1950s: The man in the chair)

In a famous McGraw-Hill ad from the late 1950s, printed in the trade press, a man sits in an office chair. He does not smile.

He speaks straight to the salesman who has just walked into his office: "I don't know who you are. I don't know your company. I don't know your company's product... Now, what was it you wanted to sell me?"

The line printed under him: **sales start before your salesman calls.** McGraw-Hill was selling ad space in its own magazines, [MarketingProfs notes](https://www.marketingprofs.com/5/nicastro4.asp), but the ad was still right.

*The short answer:* **Early, in large groups, and often before they talk to anyone.** A typical purchase now involves 13 people inside the buying company and 9 outside advisers who sway the decision, [Forrester reported](https://www.forrester.com/press-newsroom/forrester-2026-the-state-of-business-buying/) in January 2026. The internal group has grown from 8 to 13, [it added](https://www.forrester.com/blogs/predictions-2027-ai-takes-over-business-buying-but-humans-retain-control/) on 28 October 2026.

### He no longer sits alone: 13 buyers inside, 9 outside

Most of those people would rather not meet your salesperson at all. In a [Gartner survey](https://www.gartner.com/en/newsroom/press-releases/2026-03-09-gartner-sales-survey-finds-67-percent-of-b2b-buyers-prefer-a-rep-free-experience) of 646 business buyers, 67% said they prefer to buy without a sales rep. Yet in another [Gartner survey](https://www.gartner.com/en/newsroom/press-releases/2026-05-20-gartner-survey-finds-sixty-nine-percent-of-b-two-b-buyers-turn-to-sales-reps-to-validate-ai-generated-insights), published in May, 69% still want a salesperson to check what AI told them. The salesperson now comes in late, to confirm.

Think of a real deal you won or lost last year. Then be honest: how many of the 22 people had heard of you before your salesperson called?

**Chart:** One chair became 22

Thirteen people inside the buying company and nine outside it sway a typical business purchase.

- The 1950s ad: one man in the chair
- Before: 8 inside the buying company
- 2026: 13 inside, 9 outside

13 inside, 9 outside: [Forrester, January 2026](https://www.forrester.com/press-newsroom/forrester-2026-the-state-of-business-buying/). 8 to 13: [Forrester, 28 October 2026](https://www.forrester.com/blogs/predictions-2027-ai-takes-over-business-buying-but-humans-retain-control/).

#### Opinions keep moving between purchases

In October the data company 6sense shared a [preview](https://revcity.6sense.com/home/discussion/2592/counting-down-to-this-years-buyer-experience-report) of its 2026 buyer report. It says that between buying cycles, 71% of buyers change their view of which suppliers are worth considering. 6sense sells software for this problem, so read it as a vendor's number.

**Two industrial examples: Siemens and SKF**

So how do you get onto a list that is written before you arrive? Two industrial companies decided to put on a show that buyers would remember.

Siemens gave its first keynote at CES, the huge tech fair in Las Vegas, in 2024 and came back for a second in 2026. Its agency [Invision says](https://www.iv.com/case-studies/siemens-ces-2024-2026) that in 2023, 40% of surveyed US audiences saw Siemens as a tech company. By 2025 it was 87%.

#### How a bearings maker went to space

SKF makes the small steel bearings that let the parts of almost every machine turn. In 2025 it launched a "space programme" that never leaves Earth: underwater kites off the Faroe Islands that make electricity from the tides the Moon pulls. It won the Creative B2B Grand Prix at Cannes Lions, advertising's biggest awards festival, in June 2026, [LBB reports](https://lbbonline.com/news/The-Faroe-Islands-Space-Programme-One-Grand-Prix-for-SKF-One-Giant-Leap-For-Excellence). Siemens' numbers come from its agency and SKF's evidence is an award, so treat both as recipes rather than proof.

> The man in the chair did not know you. His assistant already has a view of you, right or wrong.

**If I joined your team on Monday:** I would pull your last ten deals from the CRM and ask one simple question: where did this buyer first hear of us? If nobody knows, that is the first fix.

### Three actions for your 2027 plan, from chapter 2: The list in the buyer's head

- **Action 4.** Make two CRM fields required: where each buyer first heard of you, and which rivals they considered. Then report how often you were on the buyer's first list. *What this means:* Your CRM is your sales system. The first list is the buyer's early shortlist, often written before they talk to any supplier. Count it across deals you won and lost. *Tags:* costs time, B2B, sales.
- **Action 5.** Map your last five wins: the proof finance, IT, daily users and outside advisers each needed. Fill the gaps. *What this means:* Thirteen people inside and nine outside each need a different reason to say yes. It is easy to brief only one of them. *Tags:* costs time, B2B, marketing.
- **Action 6.** Protect the brand budget, and in B2C protect it from promotions too. Agree with finance on the one number that will judge it. *What this means:* One good number is share of search: searches for your name as a share of all brand searches in your category, judged over two or three quarters. *Tags:* costs money, B2B, B2C, finance.

## Chapter 3: Will AI agents buy from us in 2027?

**The buyer that isn't human** (1916: The day shoppers stopped handing over the list)

Before 1916 an American grocery shopper handed her list to a clerk behind the counter, and the clerk walked the shelves and picked the goods for her.

In September 1916 Clarence Saunders opened Piggly Wiggly in Memphis, where shoppers took a basket and chose from the shelves themselves. He [patented](https://patents.google.com/patent/US1242872A/en) the "self-serving store" the next year. From then on the package had to do the clerk's job, [the Smithsonian's history](https://invention.si.edu/invention-stories/one-way-supermarket-aisles) of the aisle shows. **A century later, the list is going back to a clerk. This clerk is software.**

*The short answer:* **They will do the looking.** Forrester predicts that AI agents will play a part in 75% of business purchases in 2027. They will mostly search for suppliers, compare them and help with tender documents. The final step usually stays with people. [The prediction](https://www.forrester.com/press-newsroom/forresters-2027-b2b-go-to-market-predictions-ai-agents-will-influence-75-of-b2b-purchase-decisions/) came out on 21 October 2026.

### Agentic commerce: payment networks get ready

The payment networks are already getting ready for this. This summer more than 30 European banks used Visa to complete real purchases made by AI agents. The cardholder approved each one, [Visa said](https://www.visa.co.uk/about-visa/newsroom/press-releases.3457328.html) in July.

Shoppers are slower. Only 11% of American shoppers would let AI make the purchase decision, even for small things, but 31% would let it narrow the choice. That is from a [Gartner survey](https://www.gartner.com/en/newsroom/press-releases/2026-05-27-gartner-survey-finds-consumers-want-ai-shopping-help-but-not-ai-purchase-decisions) published in May 2026.

So, for 2027: agents already do the looking. For now, people still do the paying. For consumer brands, that means an agent may narrow the choice before the shopper looks. In business buying, Forrester expects AI agents to become a measurable part of purchasing.

**Illustration on the page:** Your product page, seen by a person and by an agent.

#### The odd detail in the same research

In [Forrester](https://www.forrester.com/blogs/predictions-2027-ai-takes-over-business-buying-but-humans-retain-control/)'s 2026 survey, 82% of business buyers said their purchase included some AI features, "whether they wanted it or not", as Forrester puts it. Forrester also predicts that risk reviews will delay or block a third of large companies' AI software purchases. So the same machine that shortens the search for a supplier makes the security questionnaire after it longer.

> The pack learned to sell without a clerk. Now your data has to sell to one.

**If I joined your team on Monday:** I would ask an AI agent to find a supplier like you, then sit next to it with a coffee and watch where it gets stuck on your site, and on the sites of your two closest rivals.

### Three actions for your 2027 plan, from chapter 3: The buyer that isn't human

- **Action 7.** Publish prices, specs, stock and delivery times as plain data an agent can read without logging in. *What this means:* An agent cannot ask for a quote or call a sales rep. If the number is not on the page, you are not in its comparison. *Tags:* costs time, B2B, B2C, web team.
- **Action 8.** Decide which AI agents may use your site and checkout, then tell IT, legal and customer service. *What this means:* Block, allow or verify: payment networks now offer ways to recognise agents that act for a real cardholder. Decide before a customer's agent shows up. *Tags:* costs time, B2B, B2C, IT and legal.
- **Action 9.** Build a procurement pack an agent can check: certificates, data protection terms, security answers, references. *What this means:* Forrester expects AI agents to become part of procurement, with legal, security and procurement teams asking for evidence. Answer the security questionnaire before it is sent. *Tags:* costs time, B2B, sales.

## Chapter 4: What kind of proof do buyers believe in 2027?

**Proof you can touch** (1854: The man who cut his own rope)

In 1854 a lift was only as safe as its rope. At New York's Crystal Palace, a giant glass exhibition hall, a crowd watches Elisha Otis step onto an open platform. The platform is pulled up high above their heads.

Then he orders his men to cut the rope in front of everyone. The platform drops a few inches and his safety brake holds, [Otis's own history](https://www.otis.com/en/us/our-company/history) records. **He could have printed a brochure about safe lifts. He chose to stand on one.**

*The short answer:* **Proof you cannot fake: what unpaid people say, and tests that cost you something.** When people decide whether to trust a brand, 41% put what unpaid people say about it first. Only 20% put the brand's own words first, [Edelman found](https://www.edelman.com/sites/g/files/aatuss191/files/2026-06/2026%20Trust%20and%20Brands%20-%20Top%2010%20Findings.pdf) in June 2026.

**Chart:** 41% trust what unpaid people say most, 20% what the brand says

Which voice weighs more when people decide to trust a brand

- What unpaid people say: 41%
- What the brand says: 20%

Paid voices count for less too. In Gartner's September survey, 35% of American consumers said they rely on influencers less because of AI, and 43% rely more on real people.

[Edelman Trust and Brands 2026](https://www.edelman.com/sites/g/files/aatuss191/files/2026-06/2026%20Trust%20and%20Brands%20-%20Top%2010%20Findings.pdf); [Gartner, 22 September 2026](https://www.gartner.com/en/newsroom/press-releases/2026-09-22-gartner-marketing-survuey-finds-35-percent-of-consumers-rely-on-influencers-less-due-to-ai), 1,006 US consumers.

### Buyers now check your claims with a machine

Forrester [predicts](https://www.forrester.com/blogs/predictions-2027-b2b-go-to-market-teams-must-adapt-as-ai-shifts-control-to-buyers/) that two-thirds of business suppliers will lose deals in 2027 because buyers cannot verify what they claim. Buyers already use AI tools to gather information about suppliers, sum up their claims and check the facts, Forrester notes. So an empty claim that used to cost nothing can now cost you your place on the shortlist.

**Two recent examples: safety tools and 144 bars called Paco**

#### Safety notes on the tools builders hold all day

Some of the best proof is something you can hold in your hand. In 2024 Point.P, a French chain that sells building supplies, engraved funny safety messages onto the tools builders use every day. It paid for no advertising space at all and won a Gold Effie in 2025, a prize for marketing that works. Its [own figures](https://www.effie.fr/prix/safety-tools-point-p-grave-la-prevention-dans-le-quotidien-des-artisans/) show sales of protective gear up 57% in January 2025 against January 2024.

#### 144 bars called Paco, and none had met

In Spain there are 163 bars called Pepe, 156 called Manolo and 144 called Paco, [Heineken noticed](https://lbbonline.com/news/Heineken-Neighbourhood-Bars-Tocayos-LePub). The brewer linked the bars that share a name into one network with training and support for the people who run them.

"Tocayos", Spanish for people who share a name, won a Gold Lion in Creative B2B at Cannes Lions in June 2026, [from the same jury](https://www.contagious.com/en/article/cannes-lions-2026-creative-b-2-b-winners-c76bcb63b5494f86ae94228079c19f87) that gave SKF its Grand Prix. The campaign's own case says the scheme [contributed to a 20% rise](https://www.llllitl.fr/2026/06/heineken-tocayos/) in sales at the bars that joined, compared with the year before.

> A claim costs nothing to make, and buyers know it. They believe what still costs you something.

**If I joined your team on Monday:** I would look for the proof you already have and never show: a customer who would talk on camera, a test result, a number from the service team.

### Three actions for your 2027 plan, from chapter 4: Proof you can touch

- **Action 10.** Judge your next event by sales conversations started within 90 days. Follow up every lead within two days. *What this means:* An event is proof in person. Its value shows in conversations, not in how many badges you scanned. *Tags:* costs time, B2B, sales.
- **Action 11.** Swap one brochure for proof buyers can try: a pilot, a calculator on their numbers, a sample. *What this means:* Buyers already use AI tools to check what suppliers claim, Forrester notes. A test on their own numbers holds up. *Tags:* costs money, B2B, B2C, marketing.
- **Action 12.** Film your experts and customers, and say clearly when they speak for you. Pay to promote the posts that led to meetings, but take that money from the short-term sales budget, not the brand budget. *What this means:* Saying it clearly means a visible label when an employee or customer speaks for you, or got something in return. The short-term sales budget is money meant to bring sales soon. *Tags:* costs money, B2B, B2C, marketing.

## Chapter 5: How do brands stand out when everyone uses the same AI tools?

**The sea of sameness** (1915: A bottle you could know in the dark)

On 26 April 1915 a room full of Coca-Cola bottlers voted to spend up to 500 dollars on a new bottle. They were tired of copycats in straight bottles that looked like everyone else's.

Their brief asked for a "bottle so distinct that you would recognize it by feel in the dark or lying broken on the ground", [the company's history](https://www.coca-colacompany.com/about-us/history/the-history-of-the-coca-cola-contour-bottle) records. The Root Glass Company won with a shape borrowed from an illustration of a cocoa bean. **The bottlers wanted to stop copycats and ended up with one of the most recognised shapes in the world.**

*The short answer:* **With what only they own.** In a [Gartner survey](https://www.gartner.com/en/newsroom/press-releases/2026-09-22-gartner-marketing-survuey-finds-35-percent-of-consumers-rely-on-influencers-less-due-to-ai) published in September 2026, 65% of American consumers said brands put out too much AI-made content. And 57% said it made them trust brand messages less. A shape, a colour or a sound that people already link to you is the one thing that a tool everyone uses cannot make average.

**Chart:** Could you name it in the dark?

Three bottles, no names, no logos, no colour. Just the shape.

Two shapes did the work of a logo. The third could belong to anyone.

Drawings by FAFB. Bottle 1 is traced from US design patent D48,160 (A. Samuelson, 16 November 1915), public domain. Bottle 2 is a generic glass ketchup bottle with a keystone-shaped label. Bottle 3 is made up.

No brand names or logos are shown. Bottle and label shapes may be trademarks of their owners and are shown for commentary.

**Two examples: AI Christmas ads and a fry box**

### Same tool, opposite Christmas

In 2025 Coca-Cola made its Christmas ad with AI for the second year running. It was built around the red trucks that people in Europe and America have remembered since the "Holidays Are Coming" campaign first ran in 1995. People online criticised the use of AI, but viewers liked the ad. It scored 5.9 stars, the top mark on [System1's test](https://system1group.com/ad-of-the-week/ai-or-no-ai-coke-gets-the-christmas-love), which credited "the assets Coke have developed over decades".

McDonald's in the Netherlands took its AI Christmas ad down within days after critics called it unsettling, [NBC News reported](https://www.nbcnews.com/world/europe/mcdonalds-ai-generated-christmas-advert-social-media-backlash-rcna248590). Both used the same kind of tool and got opposite results. In System1's view, the Coca-Cola ad worked because it used brand signals people already knew.

### The fry box that sells ketchup

The Heinz campaign did not need AI. The brand and its agency Rethink noticed something about the paper fry boxes in snack bars around the world. They already have the shape of the Heinz label, known as the keystone: wide at the top, narrow at the bottom. The idea won the Print and Publishing Grand Prix at Cannes Lions, [llllitl reports](https://www.llllitl.fr/2026/07/heinz-look-familiar/).

A rival can copy a list of product features in an afternoon. A look that strangers copy for you takes decades to build.

### The labels are arriving: what the EU AI Act asks of marketers

Under Article 50 of the EU AI Act, deepfakes have had to be labelled since 2 August 2026. These are AI-made images, sound or video that look real but are not. AI tools already on the market get until 2 December 2026 to mark what they make so that software can detect it, [the Commission explains](https://digital-strategy.ec.europa.eu/en/faqs/transparency-obligations-under-article-50-ai-act).

> Everyone has the same tools now. Nobody else has your "bottle".

**If I joined your team on Monday:** I would put your last ten ads next to your rivals' ads on one big table, cover all the logos with sticky notes and ask everyone in the room to point at the ones that are yours.

### Three actions for your 2027 plan, from chapter 5: The sea of sameness

- **Action 13.** Name three to five brand signals only you own, such as a colour, shape or sound, then write them into every brief, prompt and creator contract. *What this means:* A brand signal is anything people link to you alone: a colour, a shape, a sound, a character, a person. *Tags:* costs time, B2B, B2C, marketing.
- **Action 14.** Cut one series of generic content. Pay one named expert to share a monthly view backed by your data. *What this means:* Average content is now free to make and easy to ignore. A named person with your data is neither. *Tags:* costs money, B2B, B2C, marketing.
- **Action 15.** Test emotional ads on viewers before launch, not in the comments. Label AI-made content, even where no law requires it. *What this means:* The EU AI Act already requires labels on deepfakes. Labelling the rest costs little, and nobody can say later that you hid it. *Tags:* costs money, B2C, marketing.

## Chapter 6: How should brands handle price rises and green claims in 2027?

**The price of trust** (1683: The kimono shop that stopped bargaining)

In Edo, today's Tokyo, cloth merchants used to carry their goods to the homes of rich customers. The price of a kimono depended on who was asking, how well the seller knew them and how long they would take to pay. In 1683 the Mitsui family's shop Echigoya put up a new rule: *genkin kakene nashi*, cash on the counter and no inflated asking prices.

Everyone who walked through the door paid the same price for the same cloth. That is how the Mitsui family's own archive, the [Mitsui Bunko](https://mitsui-bunko.or.jp/archives/eng_story04/), tells the story, and [Mitsui & Co.](https://www.mitsui.com/jp/en/roots/1209869_7253.html) gives the same year for its cash sales at fixed prices. **Western department stores made the same promise more than a century and a half later.**

**Chart:** The price for each customer, before and after 1683

| Customer | Before 1683 | From 1683 |
| --- | --- | --- |
| A samurai | depends | same price |
| A merchant | depends | same price |
| A neighbour | depends | same price |
| A stranger | depends | same price |
| A child sent to buy | depends | same price |

*The short answer:* **In the open, and on the pack.** On 13 May 2026 a regional court in Bremen [ruled](https://www.landgericht.bremen.de/sixcms/media.php/13/PM%2035-26%20Urteil%20im%20Verfahren%20Verbraucherzentrale%20gegen%20Mondelez.pdf) that shrinking Milka bars from 100 to 90 grams in a near-identical pack misled shoppers, and Mondelez [has appealed](https://rundschau.de/produzenten/urteil-um-die-milka-tafeln-mondelez-geht-in-berufung/). Green claims need proof too: in countries that have written the EU's EmpCo rules into law, a vague claim like "eco-friendly" is banned unless the product has recognised excellent environmental performance, such as the EU Ecolabel.

**Chart:** The price per 100 grams rose from 1.49 to about 2.21 EUR, roughly 48% more

Ten grams are hard to see. The pack looks almost the same.

- Before: in the bar 100 g, usual shelf price 1.49 EUR, per 100 g 1.49 EUR
- After: in the bar 90 g, usual shelf price 1.99 EUR, per 100 g about 2.21 EUR

Landgericht Bremen, 13 May 2026: misleading. First instance, under appeal.

Weights and shelf prices: [Verbraucherzentrale Hamburg](https://www.vzhh.de/presse/milka-alpenmilch-ist-mogelpackung-des-jahres-2025). Price per 100 g is my arithmetic on those two prices. Generic bar, not the real pack.

### Shrinkflation: who reads your pack now

Larger food shops in France must put up notices on products that have shrunk, a rule in force [since July 2024](https://www.service-public.gouv.fr/particuliers/vosdroits/F34344). South Korea [fines companies](https://www.koreatimes.co.kr/www/biz/2024/05/602_373963.html) that cut sizes without saying so. And customers have less money to spare: nearly 60% of American consumers say they have less to spend this holiday season, [Gartner reported](https://www.gartner.com/en/newsroom/press-releases/2026-10-26-gartner-marketing-survey-finds-nearly-60-percent-of-us-consumers-have-less-to-spend-this-holiday-season) on 26 October.

### Green claims: the date is set, but not every country is ready

EmpCo, the EU directive on empowering consumers for the green transition ([Directive (EU) 2024/825](https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=CELEX:32024L0825)), bans vague claims like "eco-friendly" unless the trader can show recognised excellent environmental performance, such as the EU Ecolabel or an officially recognised EN ISO 14024 type I label. It also bans product claims such as "climate neutral" that rely on offsets.

Its rules apply from 27 September 2026 in the countries that have already written them into their own national law. Germany has. Eight member states, Slovenia and Spain among them, still had no national rules in late October, and France had no final law, [Mayer Brown notes](https://www.mayerbrown.com/en/insights/publications/2026/10/the-eu-directive-on-empowering-consumers-for-the-green-transition-two-years-on-what-it-requires-and-where-the-member-states-stand). There, the existing law against misleading advertising still applies.

EmpCo covers marketing to consumers, while claims made to business buyers fall under the general ban on misleading advertising. If you sell across the EU, plan as if the bans apply everywhere.

**The B2B version: you can raise the price once**

Companies that sell to companies have their own Milka. After Broadcom bought VMware, licence prices rose. In a January 2026 survey of 302 North American IT leaders, 86% said they were cutting back on VMware, [Network World reported](https://www.networkworld.com/article/4133925/some-enterprises-are-dropping-vmware-just-not-all-at-once.html). A cloud cost-management firm, CloudBolt, paid for the survey, so read it with that in mind.

When customers cannot leave easily, a price rise can bring in more money at first, as Broadcom planned. After that, customers start planning how to leave.

> Mitsui put one price on the wall. Put your price change on the pack.

**If I joined your team on Monday:** I would put every claim you make on one page and read it with sales and legal at the same table. Then we would delete a few.

### Three actions for your 2027 plan, from chapter 6: The price of trust

- **Action 16.** Explain any price rise on the pack, shelf, website and renewal letter. Offer a cheaper version, not a discount. *What this means:* A cheaper version is a smaller or simpler product under the same name, so the customer chooses. *Tags:* costs time, B2B, B2C, marketing.
- **Action 17.** List every claim you make in one place within 90 days. Drop any green claim you cannot prove. *What this means:* One list makes legal review possible, and it is the same list an assistant will check you against. *Tags:* costs time, B2B, B2C, legal.
- **Action 18.** Count people with recorded consent who clicked an email in the last 180 days. Plan on them and stop reporting open rates. *What this means:* Recorded consent means you can show when and how each person said yes. Open rates are unreliable because some email apps open messages automatically. *Tags:* costs time, B2B, B2C, marketing.

## Chapter 7: How much time will AI save a marketing team in 2027?

**The 40% promise** (1882: Why factories did not speed up for 40 years)

In 1882 Thomas Edison's Pearl Street station began selling electricity in New York. Anyone watching the lights come on would have expected American factories to get faster within a few years. They did not.

The economist [Paul David](https://ideas.repec.org/a/aea/aecrev/v80y1990i2p355-61.html) showed that American factory productivity barely moved until the 1920s. Factories had swapped the steam engine for an electric motor. But they kept the old layout: one engine turned one long spinning shaft, and that shaft drove every machine in the hall. **The gain came only when factories were rebuilt around a small motor in each machine.**

*The short answer:* **Far less than the headlines say.** American workers said generative AI saved them 2.2% of their work hours in the second quarter of 2026 (survey by Bick, Blandin and Deming, [St. Louis Fed's FRED Blog](https://fred.stlouisfed.org/blog/2026/08/does-generative-ai-save-time-at-work/)). My estimate: a team that uses AI on purpose gets back 3 to 7% of its working hours. That is time, not cash.

**Chart:** Same power, different hall

- 1900 hall, one engine, one shaft: One shaft, one speed. Machines sit where the shaft runs, not where the work flows. If the engine stops, the hall stops.
- 1925 hall, a small motor on every machine: Each machine runs at its own speed, and the hall is laid out around the work. Only then did productivity move.

Schematic, after Paul A. David, ["The Dynamo and the Computer"](https://ideas.repec.org/a/aea/aecrev/v80y1990i2p355-61.html), American Economic Review 80(2), 1990.

A century later the economist Robert Solow saw the same puzzle with computers. Offices were full of new machines that did not seem to make anyone more productive. "You can see the computer age everywhere but in the productivity statistics," he wrote in a [1987 book review](https://braddelong.substack.com/p/reading-robert-solows-review-of-cohen). The machines change long before the work does.

### How much time AI saves: 40, then about 7, then 2.2

So where does the 40% come from? It comes from single tasks measured in a lab. In a [2023 experiment in Science](https://www.science.org/doi/10.1126/science.adh2586), professional writing tasks took 40% less time with ChatGPT.

Look at a whole job and the number shrinks. Danish marketers whose employers encouraged them to use chatbots reported saving about 7% of their hours, in [survey data](https://bfi.uchicago.edu/wp-content/uploads/2025/04/BFI_WP_2025-56-3.pdf) by Anders Humlum and Emilie Vestergaard. Look at the whole economy and it shrinks again.

**Chart:** Zoom out: the 40% shrinks

One number measures a task, one a job, one an economy. Plan on getting back 3 to 7% of your hours.

- 40%: One writing task (experiment, 2023)
- about 7%: A Danish marketer's job, employer encourages AI (self-reported)
- 2.2%: All US work hours (self-reported, 2026)
- 3 to 7%: Plan on this (my estimate)

Noy and Zhang 2023; Humlum and Vestergaard, survey data to 2024; Bick, Blandin and Deming via FRED Blog, August 2026. Each grid is 100% of a different whole: one task, one job, all US work hours. The 3 to 7% is my estimate.

#### The meeting room, told by an analyst

Forrester [predicts](https://www.forrester.com/press-newsroom/forresters-2027-b2b-go-to-market-predictions-ai-agents-will-influence-75-of-b2b-purchase-decisions/) that in 2027 marketing leaders will cut up to 25% of roles because they expect productivity gains. The gains will fall short. Then they will either cut programme budgets to pay for hiring people back, or accept a smaller team that does less. That is our opening scene, with a percentage attached.

Some of it has already happened: 18% of marketing leaders [told Gartner](https://www.gartner.com/en/newsroom/press-releases/2026-09-29-gartner-predicts-ai-will-allow-the-majority-of-high-performing-marketing-teams-to-eliminate-traditional-bottom-rungs-of-the-corporate-ladder-by-2030) they had eliminated roles because of automation. Meanwhile pay and hours have not moved: Danish payroll records through 2024 show no effect, Humlum and Vestergaard [wrote for Brookings](https://www.brookings.edu/articles/still-waters-rapid-currents-early-labor-market-transformation-under-generative-ai/) in October.

#### A campaign nobody read

AI can cost a company money as well as save it. In May 2026 Starbucks Korea ran a promotion on the anniversary of the 1980 Gwangju uprising. Its slogan reminded people of a 1987 police cover-up. It led to calls for a boycott, and the chief executive was [fired](https://www.reuters.com/world/asia-pacific/starbucks-korea-head-fired-after-promotion-sparks-public-uproar-2026-05-19/).

Staff told the company's investigators they had "simply used artificial intelligence". The investigators blamed a broken approval chain: some of the people meant to approve it never opened the files, [the Korea Times reported](https://www.koreatimes.co.kr/business/companies/20260526/starbucks-korea-crisis-blamed-on-historical-ignorance-approval-chain-failure). Whatever time the tool saved was taken from the one step nobody should skip: checking.

Among marketing executives at American consumer brands, 64% expect their own use of AI to cause an incident that damages the brand within a year, [Forrester's CMO survey](https://www.forrester.com/blogs/predictions-2027-cmos-reckon-with-an-ai-reality-check/) found.

Where AI does pay off, the jobs are narrow. Salesforce let an AI agent work through 14,000 leads with low scores that nobody was following up. In a 30-day test, [it says](https://www.salesforce.com/au/agentforce/use-cases/customer-zero-for-marketing-anz/), that produced 84 sales opportunities and 12 closed deals. Salesforce sells this agent, so read it as a vendor's number. We do not know the deal sizes or what the agent cost.

#### The good campaign a test would kill

There is one more trap in the 40% chapter. A regional manager runs a four-week test. Two regions get her new campaign and two do not. But her weekly sales already swing up and down by about 12% on their own, without anybody doing anything at all.

**Calculator on the page:** The smallest jump your test can see.

Starting values: weekly sales swing 12%, weeks of test 4, regions per side 2, real effect of the campaign +10%.

- The smallest rise the test can see: 16.8%
- Real effect of the campaign: +10%

This test may miss a real gain. A campaign that worked could lose its budget.

This is the conversation I have with finance before a budget round. Action 21 puts it in writing.

In plain words: weekly swing means the usual week-to-week variation in sales, one standard deviation, as a percentage of an average week. The more your weekly sales swing, and the shorter or smaller the test, the bigger a campaign's effect must be before the test is likely to see it.

How I worked it out: 2.8 x weekly swing x the square root of (2 / (weeks x regions)). 95% confidence, 80% power. It assumes each week is independent of the others. Sales weeks rarely are, so treat the result as the minimum. Brand effects build over months and a short test cannot see them.

> The power is on. The hall has not been rebuilt yet. Budget for 40% and the two new hires disappear. By summer, the rest of the team is doing their work too.

**If I joined your team on Monday:** Before anyone touches the budget, I would ask the team to time two weekly tasks for four weeks, once with AI and once without. Real hours from your own office beat any headline you will read this year.

### Three actions for your 2027 plan, from chapter 7: The 40% promise

- **Action 19.** Budget AI as 3 to 7% of working hours back (my estimate), not cash. Count savings only when you actually decide to cut spending. *What this means:* Start at 3% if your team is just beginning and go towards 7% if it already uses AI a lot. AI tools also bill per use, so count the cost. *Tags:* costs time, B2B, B2C, finance.
- **Action 20.** Time two weekly tasks for four weeks, including checking AI output. Expand only if the hours fall and the errors do not rise. *What this means:* The checking is part of the job. The Starbucks Korea case shows what skipping it costs. If the hours do fall, rebuild the task around the tool. *Tags:* costs time, B2B, B2C, marketing.
- **Action 21.** Work out the smallest sales rise your test can detect, then agree the rules with finance before the next budget round. *What this means:* Statisticians call it the minimum detectable effect. The calculator in this chapter, on the page, shows it for your own weeks and regions. *Tags:* costs time, B2B, B2C, finance.

## Back in the meeting room

The 40% is still on the screen. Next October there will be another plan with another big number on it. This time someone at the table will raise a hand and ask, very politely, where it came from.

> "Ever tried. Ever failed. No matter. Try again. Fail again. Fail better."

The Irish writer Samuel Beckett, [*Worstward Ho*](https://www.goethe.de/prj/mis/en/lei/21891928.html), 1983. He did not mean it to cheer anyone up, which is exactly why it suits marketing so well.

Seven stories and one line from 1983. I liked the line so much that I named my company after it: **Fail Again. Fail Better.**

## Is your 2027 marketing plan ready? Seven questions

One question per chapter, three answers each. On the page, your three weakest answers pick your first three actions.

1. **Have you opened your pricing page with JavaScript switched off?** (Chapter 1: The machine that answers for you.) Answers: No, never / Once, a while ago / Yes, and the prices show. A gap here points to action 1.
2. **Does your CRM record where each buyer first heard of you?** (Chapter 2: The list in the buyer's head.) Answers: No / Sometimes, as free text / Yes, it is a required field. A gap here points to action 4.
3. **Could an AI agent find your prices, stock and delivery times without logging in?** (Chapter 3: The buyer that isn't human.) Answers: No, it is all "ask for a quote" / Some of it / Yes, all of it. A gap here points to action 7.
4. **Can a buyer try something on their own numbers before talking to sales?** (Chapter 4: Proof you can touch.) Answers: No, we send brochures / We do demos on request / Yes: a pilot, calculator or sample. A gap here points to action 11.
5. **Can you name three brand signals, such as a colour, shape or sound, that only your brand owns?** (Chapter 5: The sea of sameness.) Answers: Not really / Our logo and colour / Yes, and they are in every brief. A gap here points to action 13.
6. **Is there one list of every claim you make, green ones included?** (Chapter 6: The price of trust.) Answers: No / Partly, spread across teams / Yes, and legal has seen it. A gap here points to action 17.
7. **What AI saving is in your 2027 budget?** (Chapter 7: The 40% promise.) Answers: 40% or more / Around 20% / 3 to 7%, as time not cash. A gap here points to action 19.

## Work with me

I'm Anže Frantar. I have 12 years of experience in B2B and B2C marketing, including 8 in an agency and the last 4 as Creative Director. I have built marketing departments from scratch and led teams across Central and South-East Europe. Based in Ljubljana, working across Europe.

There are two ways to start.

### A free 30-minute call

How it starts:

1. **A 30-minute call.** Free, with me, no slides. You tell me where the plan is stuck, and we pick the two actions to take first.
2. **A first view in one working day.** What I would do first, in plain words. You owe me nothing.
3. **A fixed written quote.** Within two working days. No surprises later.

I reply within one working day with two or three possible times. Ask for the call on the [contact page](https://failagainfailbetter.com/contact/?topic=combined).

### Marketing 2027: the talk

This report as a talk for companies, conferences and leadership teams: 45 minutes on a video call, in Teams or similar, or 60 minutes live, in person. Ask about it on the [contact page](https://failagainfailbetter.com/contact/?topic=workshop).

### Roles, products and workshops

- [Fractional CMO](https://failagainfailbetter.com/fractional-cmo/) (1 to 3 days a week; six-month minimum): I lead your marketing team and your agencies, plan the year and the budget with you, and answer for the results.
- [Chief AI Officer](https://failagainfailbetter.com/chief-ai-officer/) (From one day a week; six-month minimum): I find where AI is worth the money, build the tools with your team and stay until people use them in a normal week.
- [Fractional CMO + AI Officer](https://failagainfailbetter.com/cmo-ai-officer/) (Both roles, one contract; split reviewed every quarter): One person holds both roles. The balance between marketing and AI is reviewed quarterly as your priorities move.
- [Marketing + AI Audit](https://failagainfailbetter.com/marketing-ai-audit/) (2 to 3 weeks; Full refund if nothing is actionable): A written plan that puts your marketing and AI changes in order.
- [FounderOS](https://failagainfailbetter.com/founderos/) (Built in four weeks, one to one): A personal AI assistant that remembers your work.
- [GTM OS](https://failagainfailbetter.com/gtm-os/) (In use within 30 days): A sales and marketing system on your own CRM and data.
- [AI workshop](https://failagainfailbetter.com/ai-workshops/) (One day): Your team learns AI, then builds working tools the same day.
- [Strategy workshop](https://failagainfailbetter.com/workshops/lego-serious-play/) (Half a day to two days; LEGO® SERIOUS PLAY® method): Everyone builds an answer with bricks before the discussion starts.

Write to anze@failagainfailbetter.com or find me on [LinkedIn](https://www.linkedin.com/in/anzefrantar/).

## Questions people ask about marketing in 2027

### What are the biggest marketing trends for 2027?

Seven trends stand out for 2027: AI search, buying groups, AI agents, proof, sameness, prices and green claims, and AI savings. Assistants answer instead of sending visitors, buyers write their shortlist early, AI agents do the looking, proof beats claims, AI-made sameness tires people, regulators watch packs and green claims, and AI saves less time than promised. On agents, [Forrester predicts](https://www.forrester.com/press-newsroom/forresters-2027-b2b-go-to-market-predictions-ai-agents-will-influence-75-of-b2b-purchase-decisions/) that AI agents will play a part in 75% of business purchases in 2027 (prediction, 21 October 2026).

### How will AI search change marketing in 2027?

Fewer buyers will reach your website, so what assistants say about you will matter more than where you rank. In US desktop data from Comscore, only 5.2% of ChatGPT sessions sent someone on to another website, against 31.1% of Google searches ([arXiv study](https://arxiv.org/abs/2607.07652), July 2026). A session and a search are not the same unit, but the direction is clear. Measure mentions, not clicks.

### What is GEO or AEO, and how is it different from SEO?

GEO (generative engine optimisation) and AEO (answer engine optimisation) mean getting your brand named and recommended in AI answers from ChatGPT, Perplexity, Gemini, Copilot and Google AI Mode. They build on SEO rather than replace it. Do not expect much traffic: two studies of about 100 websites each found AI assistants sent under 1% of all visits, [SparkToro reported](https://sparktoro.com/blog/is-ai-visibility-a-marketing-boondoggle/) on 22 October 2026.

### How do I check what ChatGPT says about my brand?

Ask the ten questions your buyers ask most, ten times each, in at least two assistants, and count which brands come back. One check tells you almost nothing. In January 2026 [SparkToro and Gumshoe](https://sparktoro.com/blog/new-research-ais-are-highly-inconsistent-when-recommending-brands-or-products-marketers-should-take-care-when-tracking-ai-visibility/) collected 2,961 AI answers, and any two gave the same list of brands less than once in 100 tries. Repeat the count every quarter against three rivals.

### Are ChatGPT ads available in Europe, and do they matter for 2027?

Yes, OpenAI brought ChatGPT ads to 31 European countries in August 2026. By 21 October 2026 more than 100,000 advertisers had run ads, [OpenAI said](https://openai.com/index/building-chatgpt-ad-stack/) (self-reported). The European Commission has designated ChatGPT a "very large online search engine", so the EU's strictest platform rules apply from January 2027. A small test budget makes sense, but it does not replace being named in the answers themselves.

### How do B2B buyers choose suppliers in 2027?

Early, in large groups, and often before they speak to any supplier. A typical business purchase involves 13 people inside the buying company and 9 outside advisers who sway the decision, [Forrester reported](https://www.forrester.com/press-newsroom/forrester-2026-the-state-of-business-buying/) in January 2026. In a [Gartner survey](https://www.gartner.com/en/newsroom/press-releases/2026-03-09-gartner-sales-survey-finds-67-percent-of-b2b-buyers-prefer-a-rep-free-experience) of 646 business buyers, 67% said they prefer to buy without a sales rep. So the shortlist is often written before your salesperson calls.

### What is agentic commerce, and will AI agents buy from us in 2027?

Agentic commerce means AI agents search, compare and sometimes complete purchases for a person or a company. In 2027 they will mostly do the looking: [Forrester predicts](https://www.forrester.com/press-newsroom/forresters-2027-b2b-go-to-market-predictions-ai-agents-will-influence-75-of-b2b-purchase-decisions/) AI agents will play a part in 75% of business purchases, while the final step usually stays with people (prediction, 21 October 2026). An agent cannot ask for a quote, so publish prices, stock and delivery times in plain text.

### What kind of proof do customers trust most in 2027?

Proof you cannot fake, such as what customers and other unpaid people say about you. In [Edelman's Trust and Brands report](https://www.edelman.com/sites/g/files/aatuss191/files/2026-06/2026%20Trust%20and%20Brands%20-%20Top%2010%20Findings.pdf) from June 2026, 41% said unpaid voices matter most when they decide to trust a brand, against 20% who named the brand's own words. Forrester also [predicts](https://www.forrester.com/blogs/predictions-2027-b2b-go-to-market-teams-must-adapt-as-ai-shifts-control-to-buyers/) that two-thirds of business suppliers will lose deals in 2027 because buyers cannot verify their claims.

### How can a brand stand out when everyone uses the same AI tools?

Use what only you own: a shape, a colour, a sound or a story that people already link to you. In a [Gartner survey](https://www.gartner.com/en/newsroom/press-releases/2026-09-22-gartner-marketing-survuey-finds-35-percent-of-consumers-rely-on-influencers-less-due-to-ai) published in September 2026, 65% of American consumers said brands put out too much AI-made content, and 57% said it made them trust brand messages less. AI can help with the work behind an ad. It should not replace the brand signals that make the ad yours.

### Do marketers have to label AI-generated content under the EU AI Act?

Yes, if it is a deepfake: AI-made images, sound or video that look real but are not. Under Article 50 of the EU AI Act, deepfakes have had to be labelled since 2 August 2026. AI-made text that informs the public on matters of public interest needs a label too, unless a person has reviewed it and takes editorial responsibility.

AI tools already on the market have until 2 December 2026 to mark what they make so that software can detect it, [the European Commission explains](https://digital-strategy.ec.europa.eu/en/faqs/transparency-obligations-under-article-50-ai-act). For edge cases, ask your lawyer.

### Does the EU ban on green claims apply from September 2026?

Only in countries that have already written it into national law. Directive (EU) 2024/825, known as EmpCo, bans vague claims like "eco-friendly" unless the product has recognised excellent environmental performance, such as the EU Ecolabel, and product claims like "climate neutral" that rely on offsets. Its rules apply from 27 September 2026 where transposed, as in Germany.

Several member states, Slovenia and Spain among them, had not done so by late October 2026, [Mayer Brown notes](https://www.mayerbrown.com/en/insights/publications/2026/10/the-eu-directive-on-empowering-consumers-for-the-green-transition-two-years-on-what-it-requires-and-where-the-member-states-stand). There, the existing law against misleading advertising still applies. EmpCo covers marketing to consumers; claims to business buyers fall under the general ban on misleading advertising. If you sell across the EU, plan as if the bans apply everywhere.

### How much time will AI really save a marketing team in 2027?

Far less than the 40% in the headlines. American workers said generative AI saved them 2.2% of their work hours in the second quarter of 2026 (self-reported, Bick, Blandin and Deming via the [St. Louis Fed's FRED Blog](https://fred.stlouisfed.org/blog/2026/08/does-generative-ai-save-time-at-work/), August 2026). Danish marketers who used chatbots at employers that encouraged them reported saving about 7% of their hours ([survey data](https://bfi.uchicago.edu/wp-content/uploads/2025/04/BFI_WP_2025-56-3.pdf)).

The 40% comes from single writing tasks in an experiment. The report's author, Anže Frantar, estimates that a team that uses AI on purpose gets back 3 to 7% of its hours.

### Will AI replace marketing jobs in 2027?

Probably not as fast as some budgets assume. [Forrester predicts](https://www.forrester.com/press-newsroom/forresters-2027-b2b-go-to-market-predictions-ai-agents-will-influence-75-of-b2b-purchase-decisions/) that in 2027 marketing leaders will cut up to 25% of roles expecting productivity gains, then either cut programme budgets to hire people back or accept a smaller team doing less (prediction, 21 October 2026). So far, Danish payroll records through 2024 show no effect on pay or hours, Humlum and Vestergaard [wrote for Brookings](https://www.brookings.edu/articles/still-waters-rapid-currents-early-labor-market-transformation-under-generative-ai/) in October 2026.

### Who can help us build a 2027 marketing plan that uses AI?

Your own team can start with the free actions on this page, but someone has to own the plan and the AI decisions together. The author, Anže Frantar, works from Ljubljana as a [Fractional CMO + AI Officer](https://failagainfailbetter.com/cmo-ai-officer/): one person for the marketing and the AI decisions, one to three days a week, with the split between the two reviewed every quarter.

Teams can also start with a [Marketing + AI Audit](https://failagainfailbetter.com/marketing-ai-audit/) of two to three weeks. The guarantee: Full refund if nothing is actionable. A good agency or an in-house hire can also do it.

### Can you present this report to our team?

Yes. Marketing 2027 is also a talk for companies, conferences and leadership teams. It comes in two formats: 45 minutes on a video call, in Teams or similar, or 60 minutes live, in person. To ask about it, write to me through the [contact page](https://failagainfailbetter.com/contact/?topic=workshop).

## Method and sources

Every statistic was checked against its source before publication, and every statistic in the text links to that source. Company results are marked self-reported, analyst forecasts are marked prediction, and figures from companies that sell a fix are marked vendor. The price per 100 g and the 16.8% test bar are my own arithmetic. The 3 to 7% is my estimate.

Nothing on this page is legal advice. Chapters 5 and 6 describe EU rules as they stood on 4 November 2026; check them for your market with a lawyer.

---

Canonical page: <https://failagainfailbetter.com/marketing-2027/>

Published 4 November 2026 by Anže Frantar, FAFB, d.o.o., Ljubljana. The picture credits are at the end of the page.