The brand story buyers hear first is the one AI tells.

In 2009 the writers Rob Walker and Joshua Glenn bought 100 objects from thrift stores and garage sales for $128.74 in total. They asked other writers to invent a short story for each one and listed the objects on eBay with the stories attached. The objects sold for $3,612.51 (Significant Objects, 2010). Some of the writers were well known, but the objects themselves had not changed. “Stories are such a powerful driver of emotional value that their effect on any given object’s subjective value can actually be measured objectively,” the two wrote.

Your products carry a story too, and many buyers now hear it from an AI assistant before they visit your website, if they visit at all. The assistant writes that story from whatever it can find about you, and being named in it matters less than what it says about what you make, what you stand for and why the buyer should believe you. To see how that story gets written, we put questions like this one to five AI models about three whey protein brands, one of them our client.

The buyer asks, and the assistant writes

Picture a buyer with a nearly empty tub of protein powder and a phone in their hand. They type a plain question: which whey protein can I trust? Google says its AI Mode “uses our query fan-out technique, breaking down your question into subtopics” and searches each of them (Google, May 2025), so the one question can become several, such as testing, ingredients, price and what other buyers report. OpenAI’s shopping research goes further and hands the buyer “a personalized buyer’s guide with the top products, key differences, tradeoffs” (OpenAI, November 2025). Our buyer now holds a story with a verdict at the end, written by software that has never tasted the product.

Many buyers now research this way:

And they tend to believe what they read. A European Data Protection Board taskforce wrote that ChatGPT’s outputs “are likely to be taken as factually accurate by end users … regardless of their actual accuracy” (EDPB, May 2024). Trust has limits: among Americans who have seen AI summaries in search, 53% trust them at least somewhat and only 6% trust them a lot (Pew Research Center, October 2025). If you sell to businesses, your salespeople may spend the first call confirming or correcting a story they did not write, because most buyers check.

Why a story in the answer persuades

An answer like this is likely to persuade for the same reason the thrift-store objects sold: it reaches the buyer as a story. In experiments by Melanie Green and Timothy Brock, readers who were more absorbed in a narrative came away holding more of its beliefs, and those beliefs were “generally unaffected by labeling a story as fact or as fiction” (Green and Brock, 2000). A later meta-analysis of 76 articles found that absorption in a story moves people’s beliefs and intentions toward it and leaves them with fewer critical thoughts (van Laer and colleagues, 2014). Those studies used written stories, and a buyer’s guide from an assistant is one.

A story also does more work than a feature list. When Ipsos studied 15,000 video ads, only about half tried to tell a relatable story. The ones that did were “2x more effective in changing behaviour than those that only focus on product features,” and ads built on unexpected, unconventional stories were “nearly three times more effective at driving memorability” (Ipsos, February 2026).

Buyers say as much themselves. In a Headstream survey of 2,000 UK adults, 79% thought it a good idea for brands to tell stories. If they loved a brand’s story, 55% said they were more likely to buy from it later (Headstream, 2015).

What the brand gives the assistant

Most marketing leaders would agree with Dentsu Creative’s 2024 survey of chief marketing officers, where 75% said that “every touchpoint can and must tell the brand story” (Dentsu Creative, 2024). The answer our buyer just read is one of those touchpoints, and no marketer wrote it, it was all AI. If all the assistant finds on your own pages is a nutrition panel and a list of certifications, it tends to take the rest of the story from reviews, forums and rivals’ pages.

A paragraph of jargon gives it little more to work with, and people struggle with it too. When the career site Kickresume gave 50 people the same memo in plain language or in office jargon, the plain-language readers recalled 34 accurate points an hour later and the jargon readers 12 (Kickresume, 2025). That was a company experiment without peer review. Careful readers are also the ones jargon impresses least: across four studies, the Cornell researcher Shane Littrell found that people more receptive to corporate jargon scored lower on analytic thinking and on a simulated workplace decision-making test (Littrell, 2026).

Borrowed slang can backfire as well. A team led by University of Iowa researchers found that slang in brand posts read as inauthentic and hurt sincere brands, though not brands with an exciting image (Pyrah and colleagues, 2025). A line like “next-generation clean nutrition” tells our buyer nothing they can picture, and it gives the assistant nothing it can check.

What five models told buyers about three whey brands

In October 2026 we put our buyer’s kind of question about three whey protein brands to five AI models: Google Gemini, Anthropic Claude, xAI Grok, OpenAI ChatGPT and Perplexity Sonar. We published the results as a case study, with the brand names changed. All three brands appeared in every model’s answers, yet the claims the models repeated set them apart.

Read as our buyer would read them, the claims every model repeated sound like the opening of a story: one ingredient, from small dairy farms, and 25g of protein in a 32g serving. Our client’s grass-fed claim named the same attribute without the farms, and four models repeated it. Its protein claim left out the serving size and reached two, although by weight the two products are level at about 80% and 78% protein. Our buyer had asked which brand to trust, and our client’s strongest evidence answers exactly that: independent reports on metals for tested lots. Three models passed it on, the same number that repeated a rival’s general statement about banned-substance testing, while buyer reports of stomach discomfort reached four, as many as our client’s most repeated claims. We follow the protein claim further in AI traffic is tiny. What AI says about your brand isn’t.

One category over one period cannot show why each claim traveled as it did, and answers can change within weeks. Solar Whey’s “25g protein per 30g serving” gave its serving size and still reached only four models, so the serving size alone does not explain the gap.

Write what you stand for, so a model can repeat it in its own words

The claims that reached every model told the buyer what the product is and where it comes from. That kind of detail speaks to what a company stands for, and many buyers say values weigh on what they buy. In Accenture’s survey of about 30,000 consumers, 62% said their purchasing consideration is driven by a company’s ethical values and authenticity (Accenture, 2018). IBM’s Institute for Business Value classed 40% of consumers as purpose-driven, people who choose products and brands that match their values (IBM, 2020).

The assistant brings views of its own to the question, based on the available information. When marketing researchers compared language models’ brand perceptions with human surveys, the two agreed more than 75% of the time (Li and colleagues, 2024). Another team found that language models disproportionately associate global brands with positive attributes (Kamruzzaman and colleagues, 2024), which can leave a local brand with more to prove. The shortlist moves as well: a September 2026 study of six models found that “repeated responses to the same query can produce different brands and rankings,” and that a buyer’s goals and constraints change which brands appear (Malthouse and colleagues, 2026). A buyer who asks for whey from small farms, or for a brand that publishes its test results, is asking about values, and the model matches that request against what it can read about each brand.

Walker and Glenn chose the writers who told their objects’ stories. With an assistant, the closest you come is choosing what your own pages give it to read, and the metals reports show what is at stake. A value written as an adjective gives the model little to repeat, while a value written as a practice, with its proof on the same page, gives it a sentence it can quote. Write those sentences the way you would explain them to a customer across a counter, and put each one on the page that carries its proof. The right-hand column comes from claims we tracked; the last row is the sentence we proposed for our client:

Questions to ask a PR firm or narrative intelligence platform

If you want help with this, agencies and software platforms offer to track and shape what AI says about brands, under names such as generative AI storytelling and narrative intelligence. Five questions show whether a firm reads the story or counts the mentions:

1. Do you read the full text of every answer, or count how often the brand is named?
2. Do you separate facts, claims and allegations, and check each one against the brand’s own published record?
3. Do you ask each major assistant each question many times, since answers change from run to run?
4. Do you deliver the sentences to publish and the page each belongs on?
5. How do you protect confidentiality, and can you work through counsel or an agency?

We read these answers for brands and for the people who lead them, and we give clients the sentences to publish. The first consultation is free and includes a sample reading of your own website. Tell us what AI should say about you.

Sources

1. Joshua Glenn, Significant Objects, Announcing the SO Challenge: Spot the Patterns!, March 6, 2010.
2. Joshua Glenn and Rob Walker, About Significant Objects.
3. Elizabeth Reid, Google, AI in Search: Going beyond information to intelligence, May 20, 2025.
4. OpenAI, Introducing shopping research in ChatGPT, November 24, 2025.
5. Athena Chapekis and Anna Lieb, Pew Research Center, Google users are less likely to click on links when an AI summary appears in the results, July 22, 2025.
6. Natasha Sommerfeld, Megan McCurry and Doug Harrington, Bain & Company, Goodbye Clicks, Hello AI: Zero-Click Search Redefines Marketing, February 19, 2025.
7. Gartner, Gartner Survey Finds 69% of B2B Buyers Turn to Sales Reps to Validate AI-Generated Insights, May 20, 2026.
8. Barbara Winters, Forrester, The State Of Business Buying: Risk-Averse Buyers Demand Proof, Not Promises, January 21, 2026.
9. European Data Protection Board, Report of the work undertaken by the ChatGPT Taskforce, May 23, 2024.
10. Kirsten Eddy, Pew Research Center, Americans have mixed feelings about AI summaries in search results, October 1, 2025.
11. Melanie C. Green and Timothy C. Brock, The role of transportation in the persuasiveness of public narratives, Journal of Personality and Social Psychology, 2000.
12. Tom van Laer, Ko de Ruyter, Luca M. Visconti and Martin Wetzels, The Extended Transportation-Imagery Model: A Meta-Analysis of the Antecedents and Consequences of Consumers’ Narrative Transportation, Journal of Consumer Research, 2014.
13. Ipsos, Misfits Stories: The Power of Storytelling in Creative Effectiveness, February 23, 2026.
14. Headstream, The power of brand storytelling, 2015.
15. Dentsu Creative, CMO Survey 2024, 2024.
16. Kickresume, Office Jargon Can Cost You Time … And $6 Million a Year!, April 10, 2025.
17. Shane Littrell, The Corporate Bullshit Receptivity Scale: Development, validation, and associations with workplace outcomes, Personality and Individual Differences, 2026.
18. Bryce Pyrah, Jing Wang, Yiyi Li and Ying Xie, The Slang Paradox: Connecting or Disconnecting with Consumers?, Journal of Marketing Research, 2025.
19. Accenture, Majority of Consumers Buying From Companies That Take A Stand on Issues They Care About and Ditching Those That Don’t, Accenture Study Finds, December 5, 2018.
20. IBM Institute for Business Value, Meet the 2020 consumers driving change, 2020.
21. Peiyao Li, Noah Castelo, Zsolt Katona and Miklos Sarvary, Determining the Validity of Large Language Models for Automated Perceptual Analysis, Marketing Science, 2024.
22. Mahammed Kamruzzaman, Hieu Minh Nguyen and Gene Louis Kim, “Global is Good, Local is Bad?”: Understanding Brand Bias in LLMs, EMNLP 2024.
23. Edward Malthouse, Kun-Yu Lee, Jing Yang, Sanchary Pal and Xueyan Feng, Evaluating Brand Retrieval and Ranking in Large Language Model Recommendations, September 14, 2026.