AI chatbot brand recommendation audit: who does it and what it checks

What an AI brand recommendation audit is

It's a check of how AI assistants talk about your brand when customers ask them about your niche. Unlike an SEO audit, the focus isn't the site itself but the answers: ChatGPT, Gemini, Google AI Overviews and AI Mode. The audit answers four questions: are you mentioned, are you recommended, what is said about you and why.

It is the starting point for any GEO work: without a baseline measurement you can't tell whether anything has changed.

Who does this kind of audit

• GEO agencies. They specialise in AI visibility and usually have their own tools and methodology. The upside is experience in this particular task. The downside is often a higher price and less knowledge of your market.
• SEO and marketing agencies. They have added AI audits to their SEO services. Convenient if you already work with them: they know your site. It's worth checking how exactly they collect AI answers.
• In-house marketers with tools. The cheapest option if someone on the team has the time. You need a service that regularly puts queries to AI assistants and stores the answers. An overview of such services is in the article on AI visibility tools.

What the audit checks

1. Mention rate and share of voice

Take 10–50 queries that customers actually ask, in their language and with their city in mind. For each, record:

• mention rate — the share of answers that mention the brand;
• recommendation rate — the share of answers in which AI actually recommends the brand;
• share of voice — your share of mentions among all brands in the answers;
• competitors that AI names instead of you.

Answers change from day to day, so a one-off check is unreliable. An honest audit is based on several days of data collection, not on a single screenshot.

2. Factual accuracy

AI may quote the wrong prices, an old address, a closed branch or a service you don't offer. The audit records every such error and looks for where it came from: often an outdated directory or someone else's article. How to fix them is covered in the article how to fix AI errors about your brand.

3. Sources and their weight

Which sites do answers in your niche cite: your site, media, listicles, maps, reviews, competitors' sites. In our monitoring, among AI answers with links, the brand's own site was cited in 54%, media, blogs and other sites also in 54%, competitors' sites in 36%. Detailed data is in the article on which sources AI relies on. The audit shows which of these sources affect your queries specifically and where you are missing.

4. Technical readiness of the site

• AI search crawler access. For ChatGPT, the key crawler is OAI-SearchBot: OpenAI says that sites that block it will not be shown in ChatGPT search answers. GPTBot is a separate crawler for model training, and its settings are independent.
• robots.txt for AI bots. Check that the crawlers you need haven't been blocked by accident. For Google AI Overviews and AI Mode, control is through the usual Googlebot rules. The Google-Extended token governs the use of content for training Gemini and, according to Google's documentation, does not affect a site's inclusion in Google Search.
• llms.txt. This is a proposal by Jeremy Howard from 2024 — a Markdown file with a short description of a site for language models. It is not an official standard. In its guide to optimising for generative AI, Google says such a file neither helps nor harms visibility in Google Search. An audit may note whether it exists, but it isn't a priority.
• Indexing. For Google's AI features, a page must be indexed and eligible to be shown with a snippet.
• Structured data. Google doesn't require special markup for AI, but it recommends continuing to use structured data as part of SEO. The audit checks that the markup matches the visible text.
• Facts as text. Prices, addresses and opening hours should be on the page as text, not only in an image.

What you should receive

1. A list of queries and platforms the check covered.
2. Real AI answers — not just numbers, but the texts and links themselves.
3. Metrics for mentions, recommendations and share of voice compared with competitors.
4. A list of factual errors with the sources they come from.
5. A source map: where you are cited, where competitors are cited, where you are absent.
6. Technical notes on the site.
7. A prioritised action plan: what to do first and why.

If an agency promises that after the audit AI is guaranteed to start recommending you, that's a reason for doubt. The answer is formed by the model, and nobody fully controls it.

How often to run an audit

A full audit with a technical check and a source map is enough once a quarter or after noticeable changes: a redesign, new services, a move. But mention and recommendation metrics are better tracked continuously: answers change every day, and without regular measurement you can't see whether your actions are working.

One option is AI Visibility from IWIS. Every day the service checks ChatGPT, Gemini, Google AI Overviews and AI Mode in your customers' language and from their city, shows mentions, recommendations, sources and errors, and builds an action plan backed by evidence. The IWIS team can implement the plan at a separate price. Details are on the pricing page.

How to check for your business

Start with a free live check: enter your brand and a customer question, and in 20–60 seconds you'll see a real ChatGPT answer. For a fuller picture across four platforms, use the free first week — in effect, a baseline audit on your own queries.