AI will tell your story anyway. Make sure it is in your own words.
AI reputation management, beyond AI visibility.
When a buyer, investor or journalist asks an AI assistant about you, it answers with a story about what you make and whether you can be trusted. We read those stories against your own website and give you the sentences that make the assistant tell yours.
Specialists in PR, communications and search, backed by our own narrative intelligence platform.
See the whey studyCase study: AI take on whey protein recommendations
We asked five AI models about three whey protein brands, the way a buyer would. Two of our client’s claims show what the models chose to repeat.
Protein per serving
All five models repeated a rival’s protein claim. Only two repeated our client’s, although by weight the two products are level, about 80% and 78% protein. The rival’s page said how big its serving was and our client’s did not, so a buyer reading the answers could not compare the two.
28g protein per labeled serving
Good Whey Labs, our client2 of 5 models25g protein per 32g serving
Energize Isolate5 of 5 models
What we proposed
“28g of protein in a 34.9g serving, about 80% protein by weight.”
Independent lab reports
Our client’s strongest evidence went no further than a rival’s one-line claim. Its independent lab reports on metals for tested lots reached three of the five models, and so did the rival’s less specific statement that its product is tested for banned substances. The reports were public, but no page set them out lot by lot.
- Independently verified for metalsGood Whey Labs, our client3 of 5 models
- Banned-substance testedEnergize Isolate3 of 5 models
What we proposed
A testing page that lists each tested lot, its flavor and its metals results against specification.
From a real study of five AI models, October 2026. Brand names and identifying details changed.
See the whey studyA complaint can travel further than your proof.
In the same study, buyer reports of stomach discomfort reached four of the five models, one more than our client’s lab reports, although no cause has been established. AI reads your pages alongside reviews, forums and resellers, and it tends to repeat whatever is easiest to find and quote. Your proof has to be at least as easy to find as the complaint.
Where else the story goes wrong
Four illustrative situations from the kinds of matters we take on.
Founders with a thin public record
A Series A+ founder is described as a senior engineer at her previous employer, a job she left four years ago. Two models add a board seat that belongs to someone else with the same name.
Products described by the wrong attribute
A cookware maker’s pans are valued by chefs for even heat across the base. AI answers describe them mainly by their color range, taken from a retailer’s listing, and recommend another brand for even heating.
Growth companies with a history
A payments startup had a two-day outage in its first year and published a full post-mortem. Answers to “Is it reliable for payroll?” still lead with the outage and never mention the redundancy added since.
Leadership changes, launches and deals
A consumer brand withdraws a new product line after complaints, and its chief executive leaves a few months later. A year on, the line has been relaunched under new leadership, yet AI answers still open with the withdrawal and the departure. Customers, investors and a would-be acquirer all hear the old story before the new one.
Buyers compare products one requirement at a time.
Buyers ask for undyed merino with no synthetic blend, or a card with no foreign exchange fee. In the whey study, grass-fed sourcing set our client apart from one rival and matched the other, so each comparison needed its own claim.
Competitors are already rewriting their pages.
In client work we have seen competitors revise their wording once or several times a week. They do not announce it.
Your website is the source of truth.
There is no second database to build or keep in sync. We read what AI says against your pages, sentence by sentence. Each finding names the pages the answer drew on, including ones you do not own.
A figure can be right while the sentence around it is wrong.
We check every figure with its unit, its conditions and its date, since a model repeats the whole sentence.
The bar for your brand
Ask an AI assistant the question your best customer asks before buying, then read the answer next to your own site.
- Present.
- You are named for the requirement you meet best, and that requirement is the reason given.
- Accurate.
- Every figure arrives with the condition you published beside it.
- Current.
- No answer describes a product, price or title you have retired.
- In your own words.
- Your main difference is described as you state it and traced to your page.
An account that meets all four tells your story the way you would. The report starts wherever one fails.
Why AI reputation management matters
AI assistants build their account of a company from the documents they reach first, often third-party pages that lag behind or leave things out, and rarely from the company’s own material. We do this work so that the story buyers hear is one the company can stand behind, told in its own words.
How an engagement runs
- A free consultation.
- We agree what the work should achieve and show you a first reading of your website.
- A written proposal.
- Scope, timing and fee, settled before any work starts.
- The analysis.
- Your buyers’ questions put to five AI models and read against your own pages. First report within 3 days of the agreed brief.
- The correction.
- The wording and pages to publish, worked through with your team or agency.
Start with a free consultation
We agree the scope, depth and outcome you want, and show you a sample of our work on your own website.
Request a free consultation