Type your brand name into ChatGPT, Gemini or Perplexity followed by one word: "reliable?" What comes back is your reputation as the machines have compiled it, and a growing share of your potential customers now sees that answer before, or instead of, your website. Reputation management has quietly acquired a new surface, with new rules. Here is what actually determines what the machine says about you, and what you can do about it.
How the machine builds its answer
When an assistant answers "is [brand] legit?", it is not consulting a secret database. It synthesises from what it can read: your review profiles and their visible patterns, discussion threads, press mentions, complaint sites, and your own pages. Three properties of that synthesis matter enormously for brands:
- It reads like an investigator, not a fan. The machine does what careful buyers do, at scale: it weighs the review record heavily, notices unanswered complaints, and reports patterns ("customers frequently mention slow shipping") rather than averages. The behaviour we described in why buyers read your worst reviews first is now automated and confident.
- It states things flatly. A human browsing your profile forms an impression; the machine issues a verdict, in complete sentences, with no body language of uncertainty. "The brand has mixed reviews, with recurring complaints about refunds" reads as settled fact to the person asking.
- It is slow to update. The answer blends sources with different freshness. A brand that fixed its operations last quarter can keep receiving last year's verdict for a while, an AI-flavoured version of the frozen reputation problem: old complaints with no fresh counter-signal stay load-bearing.
What moves the machine's answer
You cannot edit the answer. You can edit the evidence it is built from, and the levers turn out to be the ones this blog keeps returning to, with their weights rearranged:
- Fresh review volume matters even more. Recency is the machine's main defence against stale verdicts. A steady flow of recent authentic reviews is the strongest signal that the current experience differs from the old complaints. The collection system from the Trustpilot guide is now also AI-answer maintenance.
- Your replies are part of the corpus. Machines read reply threads. A complaint with a calm, specific brand response underneath gets summarised differently from the same complaint alone: "the company responds to issues" is a sentence that appears in AI answers about brands with good reply coverage, and it is earned one reply at a time.
- Third-party mentions carry disproportionate weight. Machines trust independent sources over brand pages, heavily. A single detailed mention in a press piece, a comparison article or an industry roundup often outweighs pages of your own copy. This is why being cited in others' content has become the modern version of link building.
- Your own pages still set the factual frame. Clear shipping, returns and contact information on your site gives the machine safe facts to include. Vague pages leave the complaint threads to fill the vacuum.
The audit you should run this week
Twenty minutes, three assistants (ChatGPT, Gemini, Perplexity), five questions each: "is [brand] legit?", "[brand] reviews", "[brand] shipping problems?", "[brand] vs [your main rival]", "should I buy from [brand]?" Record the answers. You are looking for: which sources it cites, which complaints it repeats, how old its picture is, and whether your replies and fixes appear at all. That transcript is your machine-readable reputation, and most founders have never read it.
Then note the asymmetry: your rivals' transcripts are equally readable. The comparison question ("X vs Y, who is more reliable?") is being asked by real buyers today, and the answer is being assembled from exactly the review hygiene this blog covers. In our study of 578 struggling brands, 56% never answer complaints; their AI verdicts are being written entirely by their angriest customers, and their competitors only need to be visibly responsive to win the comparison by default.
Your review record has become an API that other people's software queries. If you want to know what it currently returns, and how to change it, the free reputation audit now includes exactly this check.
FAQ
Can I submit corrections to AI assistants about my brand?
No meaningful correction channel exists. The route is upstream: change the review record, the replies and the third-party mentions the answers are built from.
How fast do AI answers update after things improve?
Unevenly: weeks to months depending on the assistant and source mix. Fresh review volume is the accelerator you control most directly.
Do AI assistants read review replies?
They read what is public, and reply threads are public. Responsive brands get described as responsive; that sentence in an AI answer is earned through reply coverage.
Is this replacing traditional review checking?
Supplementing, and growing. The same underlying record feeds both the human check and the machine one, which is why the fix is one system rather than two strategies.