Contact

WritingGuide

How AI answers pick which brands to name, and how we track it weekly

By · Published · Updated

Sources checked on

How this was made. AI agents drafted this piece, a separate checker agent checked its sources and claims, and Rues read and approved it. How we work with AI

Ask ChatGPT which accounting tool a five-person shop should use, and you get a short list of names with a line or two on each. Ask Perplexity or Google's AI Mode and the list may differ. For the people who make those tools, the first question is simple: why those names and not ours?

Below: what the companies behind these answers say in their own documentation about how an answer is built, what that means for a brand, and the five-step method we use at Rueseo, with the weekly table we keep.

Where do the brand names in an AI answer come from?

An answer can draw on what the model learned in training and on pages it retrieves when the question is asked. Nobody outside these companies can see how the two are weighed. The retrieval part is documented, and it is the part a brand can work on.

  • Google AI Overviews and AI Mode. Google writes that both "may use a 'query fan-out' technique", issuing "multiple related searches across subtopics and data sources" to develop a response, and then show links to supporting pages. To be eligible as a supporting link, a page "must be indexed and eligible to be shown in Google Search with a snippet".
  • Gemini. The Gemini API documentation for grounding with Google Search describes the steps: the model analyzes the prompt, decides whether a Google Search can improve the answer, and if needed "generates one or multiple search queries and executes them". The response comes back with citations. In the Gemini app, Google says responses may include a Sources button with links, though not every response has one.
  • ChatGPT. OpenAI's documentation says that when ChatGPT uses web search, search results and citations appear in the chat. Its crawler page says OAI-SearchBot is used "to surface websites in search results in ChatGPT's search features", and that sites opted out of it "will not be shown in ChatGPT search answers, though can still appear as navigational links".
  • Perplexity. Perplexity writes that PerplexityBot is designed "to surface and link websites in search results", and that when a user asks a question, Perplexity-User "might visit a web page to help provide an accurate answer and include a link to the page in its response".
  • Microsoft Copilot. Bing Webmaster Tools added an AI Performance report (public preview) on 10 February 2026. Microsoft says it shows how often a site's content is cited in Microsoft Copilot, AI-generated summaries in Bing and select partner integrations. It also lists "grounding queries", which Microsoft describes as "the key phrases the AI used when retrieving content".

None of these documents gives a rule for which brand gets named. What they do describe is a search step before the answer is written. So when a buying question triggers a search, the brand list is shaped by the pages that search finds and by what those pages say.

Which pages does an answer read for a "which tool should I pick" question?

Take a question such as "Which project management tool is best for a 10-person design studio?" The pages that can answer it fall into a few groups:

  • Comparisons, such as "Tool A vs Tool B" pages, written by the brands themselves or by third parties.
  • Reviews on review sites, in trade press and on personal blogs.
  • Lists such as "best project management tools for agencies".
  • Guides that explain how to choose, and name tools as examples along the way.
  • Brand sites: pricing pages, feature pages, help centers and case pages.

When we log the citations under a buying question, we tag each cited page with one of these groups. Two problems show up quickly. If your brand is missing from the third-party pages that keep getting cited, a search-based answer has little reason to name you. If those pages describe you with old prices or a feature you dropped, the answer can repeat the old information. Both problems sit on pages you can find, read and act on.

How to get your brand mentioned in ChatGPT answers: the method

Our method has five steps. The first two are about your own site, the next two about the rest of the web, and the last one is the measurement that tells you whether the first four worked.

1. Access: can the search bots reach your pages?

A page an engine's search crawler may not read has little chance of being cited there; OpenAI says sites opted out of OAI-SearchBot are not shown in ChatGPT search answers. OpenAI and Perplexity recommend allowing their search crawlers (OAI-SearchBot, PerplexityBot). Anthropic writes that blocking Claude-SearchBot may reduce your site's visibility in user search results. Google's guide for AI features lists "ensuring that crawling is allowed in robots.txt, and by any CDN or hosting infrastructure" as an SEO basic that still applies. Microsoft says Bing respects the preferences site owners set in robots.txt.

The CDN part is easy to miss: a CDN setting can block a bot that robots.txt allows, and nothing in the robots.txt file shows it. Our post on robots.txt for AI bots lists each bot, what its owner says it does, and the Cloudflare settings to check.

2. Answer blocks and structured data

The pages you control should answer the questions buyers ask in a form that can be quoted. On every important page we write:

  • a short answer of two or three sentences near the top;
  • a plain definition of what the product is and who it is for, and who it is not for;
  • a table where a comparison is natural: plans, limits, prices with the date they were checked;
  • a short FAQ built from questions customers actually ask.

Microsoft says this in its AI Performance announcement: "Clear headings, tables, and FAQ sections help surface key information and make content easier for AI systems to reference accurately." It also recommends keeping content fresh and using IndexNow to tell search engines when a page changes.

Structured data needs a careful note. Google writes that there is "no special schema.org structured data that you need to add" for AI features, and that you do not need "new machine readable files, AI text files, or markup" to appear in them. Its advice is to make sure structured data "matches the visible text on the page". So we use JSON-LD to describe what is already visible on the page, and we keep the two in sync. We add an llms.txt file too, and we tell clients plainly that its effect is unproven.

If you use AI tools to write these pages, every fact still needs a human check before it goes live. Google asks for exactly that in its guide on generative AI content, which we summarized in Google's guide to generative AI content.

3. Find the pages that are cited today

Before writing anything new, we run the question set from step 5, open every citation and record:

  • the address of each cited page and which group it belongs to;
  • whether your brand appears on that page, and how it is described;
  • whether a competitor appears where you do not.

Two free reports help with your own pages. Bing's AI Performance report lists which of your URLs were cited and a sample of the grounding queries behind them. Google counts traffic from AI Overviews and AI Mode in Search Console's Performance report, under the "Web" search type, together with the rest of your search traffic.

The result is two lists. One is the third-party pages that shape the answers in your category, which becomes the work list for step 4. The other is your own pages that are cited, or that should be and are not, which becomes the work list for step 2.

4. Honest mentions on other sites

The goal is that the pages from step 3 describe your brand correctly. The ways there are ordinary and slow:

  • write to the author of a list or comparison with corrected facts and a source for each one;
  • offer reviewers access to the product, with no conditions on what they write;
  • answer questions in communities and forums under your own name, with your connection to the brand stated;
  • publish original information that other people want to cite, such as clear pricing pages or your own data with its method.

What we do not do, and what we turn down when asked: fake reviews, reviews posted by staff as if they were customers, payment for coverage that hides the payment, and paid links passed off as editorial ones. In the United States, the Federal Trade Commission's final rule on reviews and testimonials, announced on 14 August 2024, prohibits creating or selling fake reviews, paying for reviews on the condition that they say something positive or negative, insider reviews that do not disclose the writer's connection to the business, and company-controlled review sites that claim to be independent. Google's spam policies count buying links for ranking as link spam, and say paid links are acceptable when they carry rel="sponsored" or rel="nofollow". If a placement is paid, we label it as paid. We apply the same line in every country we work in.

5. Measure every week with the same questions

Google writes that AI Mode and AI Overviews "may use different models and techniques, so the set of responses and links they show will vary". That is two surfaces from one company. We do not treat one check on one day as a measurement; to see a trend, the inputs have to stay fixed.

How we set it up:

  1. We write a fixed set of questions in the brand's buyers' own words, split by intent: category questions ("best X for Y"), comparisons ("A or B"), problem questions ("how do I fix Z") and brand questions ("is A any good").
  2. Each question goes to ChatGPT, Gemini, Perplexity and Copilot.
  3. We run them in a fresh chat, in the same language and country setting, and record the date of each run.
  4. For each answer we record whether the brand is named, where it sits in the list, whether the description is correct, and which pages are cited.

Here is what one week of that sheet looks like. This is an example only: "Brand A" and every value below are made up to show the columns.

WeekQuestionEngineBrand A named?Place in listDescription correct?Brand A page citedOther pages cited
2026-W41Best invoicing tool for freelancersChatGPTYes3 of 5No, old priceNone2 lists, 1 review
2026-W41Best invoicing tool for freelancersPerplexityYes2 of 4YesPricing page1 comparison, 1 guide
2026-W41Brand A or Brand B for a small agencyCopilotYes1 of 2YesFeature page1 review
2026-W41Best invoicing tool for freelancersGeminiNoNoneNot namedNone2 lists, 1 forum thread

Each week we report three numbers next to the starting value: the share of answers that name the brand, the share of those that describe it correctly, and how many of the brand's own pages were cited. The last column feeds step 3, so next week's work list comes out of the measurement.

The limits go on the sheet too. A fresh chat shows one sample of what users may see. Some answers cite nothing. A question set measures only its own questions.

What this method cannot promise

No company in this post offers a way to guarantee a mention, and we do not offer one either. Google says that even a page meeting every requirement may not be crawled, indexed or served. Changes also take time: OpenAI says a robots.txt change takes about 24 hours to reach its search systems, and pages on other sites have to be crawled again before an answer can use them. What you do get is a dated record, every week, of where you stand on the questions you care about and which pages are behind each answer.

Where to start this week

  1. Write down ten questions a buyer would ask before choosing a product like yours.
  2. Ask them in two or three of the engines above and save the cited pages.
  3. Check that your robots.txt and CDN let the search bots in.
  4. Read the three most cited third-party pages and note how they describe you.

For step 3, Ruescan, our free site check, reads your robots.txt for AI bot access and also checks your sitemap, llms.txt, structured data, meta tags, speed and security, with no sign-up. If you want the full method run for your brand, with the weekly table, write to us using the form below. More on the service: SEO, GEO and AEO.

Frequently asked questions

Can I pay to be mentioned in an AI answer?

Not into the answer text. None of the documentation cited here describes a way to buy a place in what an answer says. Ads are a separate channel: Google's Ads help page says ads can show above or below AI Overviews, and in some countries, in English, inside them. They are chosen through the usual ad auction from Search, Shopping and Performance Max campaigns, and advertisers cannot target the spot inside an AI Overview directly. Sponsored content on other sites is also possible, but it has to be labeled as paid, and under Google's spam policies its links should carry rel="sponsored" or rel="nofollow".

Do I need llms.txt or special schema to appear in Google's AI Overviews?

No. Google says there are no additional requirements and no special schema.org markup for AI Overviews or AI Mode, and that you do not need AI text files. A page needs to be indexed and eligible to appear in Google Search with a snippet.

Why do ChatGPT and Gemini give different lists for the same question?

They are separate products from separate companies, and each runs its own search step and its own model. Even within Google, AI Mode and AI Overviews may use different models and show different responses and links. That is why we compare each engine only with its own earlier weeks.

Is Search Console enough to measure AI visibility?

It is a useful part. Google counts AI Overviews and AI Mode traffic in the Performance report together with the rest of web search, and Bing's AI Performance report counts citations. Neither tells you whether your brand was named in an answer that did not link to you, which is why we keep a question set as well.

Sources

Want to talk about this for your brand?

Write about your project and what you want to do. The message goes straight to Rues.

Open the contact form

Emailinfo@ruesandora.com Telegramruesandora X@ruesandora0