Why ChatGPT and Perplexity do not mention your brand, and how to measure it

How do you measure whether ChatGPT or Perplexity mentions your brand?

Ask a fixed panel of real questions that do not contain your brand name, on each engine you care about, several times, in fresh chats, and record who is named and which sources are cited. One answer proves nothing: the same question can name different brands from one run to the next.

1. Write 15 to 30 questions per market, in the words your buyers use: discovery ("which tool for X?"), comparison ("A or B?"), problem ("why does X happen?") and decision ("is X worth paying for?"). Keep the few questions that contain your name in a separate list: they test accuracy, not visibility.
2. Freeze the panel. If you reword a question, this month's result can no longer be compared with next month's.
3. Use a private window or a temporary chat, logged out or with memory off, web search on, and one country for the whole test. Write down the engine and the mode it shows.
4. Open a new chat for every question, paste it exactly, and copy the full answer and the sources shown with it.
5. Run each question three times per engine, ideally on different days. Report presence as the share of runs that name you, not as yes or no.
6. Run the same panel for your two or three main competitors' names too. If everyone moves at once, the engine changed, not you.

• Name Drop: a skill in SEO Skills by PENRA that builds this panel, runs it on Claude's own web search, takes the answers you paste from ChatGPT, Perplexity or Gemini, and scores the result out of 100.

What should you write down for each AI answer?

For each run, record whether you are named, where, who is named instead, which domains are cited as sources, and what is said about you. The cited domains matter most: they are the pages the engine read before answering, so they tell you where your name is missing.

A mention is your name in the text of the answer. A citation is a link to one of your pages among the sources. You can get one without the other, so count them separately.

Your own data completes the panel: visits referred by chatgpt.com, perplexity.ai or gemini.google.com in your analytics, and searches for your brand in Search Console. Google counts sites shown in AI Overviews and AI Mode in the overall search traffic of the Performance report, under the "Web" search type, not as a separate line.

What did this method find on our own four-day-old site?

We ran this panel method on seoskills.click on 15 October 2026, four days after launch: the brand was named in 0 of 34 answers to questions without its name, in English and French. It was one search-based engine (Claude's own web search) with one run per question, so it is a baseline, not a verdict on ChatGPT or Perplexity.

The panel had 40 questions: 34 without the brand name and 6 with it. Even asked by name, the engine identified the right product in 0 of 6 runs. Three runs found nothing, two returned career articles about "SEO skills", and one answered "is seoskills.click legit?" with reviews of a different, similarly named service.

The 34 answers cited 232 sources: about 31% were skill directories that mostly list public GitHub repositories, 29% vendor blogs, 22% independent articles and roundups, 6% code hosts and 4% LinkedIn. One LinkedIn post announcing a free pack was named in 6 of the 34 answers. Our pack is paid and has no public repository, so it had no way into the largest group of sources the engine read.

Six questions, this page's own question among them, were answered without naming any product. Those open questions are where a new site can start.

What the audit recommends, in that order: answer-first guides like this one for the open questions, a page in English that says plainly who publishes the pack, factual pitches to the roundups the engine already cites, posts under the full name "SEO Skills by PENRA", and the same panel again in late November 2026 on ChatGPT, Perplexity and Gemini, three runs each. None of this guarantees a mention.

Why does ChatGPT not mention a brand?

For one of four reasons, most likely first for a young brand: the pages the engines cite do not name you, the engines cannot tell which entity your name refers to, their search crawlers cannot reach your pages, or your pages do not answer the question asked. Your panel tells you which applies.

1. Not in the sources. If the domains in your "sources cited" column never mention you, the engine has nothing to name. This is the usual state of a young site, not a technical fault.
2. Entity unclear or name collision. Search your exact name on its own. If the first results are another company, a generic phrase or a lookalike domain, engines will confuse you with it or skip you. Asking "What is [your brand]?" shows whether the engine resolves you at all.
3. Crawler access. Read your robots.txt and check whether the search agents (OAI-SearchBot, Claude-SearchBot, PerplexityBot, Googlebot) are disallowed, by name or by a catch-all rule. Then check your firewall or CDN bot settings and logs for 403 or 429 responses to those agents: an open robots.txt proves nothing if the firewall blocks them.
4. Content that does not answer. When a question is answered by pages that state the answer directly, check whether any page of yours states it in its first lines, with your scope, price and limits in plain text.

With web search off, an assistant answers from what its model learned in training. A brand launched after that data was collected cannot be in it, whatever you change this month: test with search on first.

Which AI crawlers should robots.txt allow, and which only collect training data?

Each company runs separate agents for search, for training and for fetching a page a user asked about. To be eligible for AI search answers, allow the search agents. Blocking a training agent is a separate choice; the exception is Google-Extended, which also governs grounding in Gemini Apps.

# Search agents allowed, training refused (a choice, not a requirement)
User-agent: GPTBot
Disallow: /

User-agent: ClaudeBot
Disallow: /

User-agent: *
Allow: /

OpenAI says a robots.txt change takes about 24 hours to be reflected, Perplexity up to 24 hours. Both publish their crawlers' IP ranges and recommend allowing requests from them: a firewall rule can undo an open robots.txt.

Google has no separate token for AI Overviews or AI Mode: a page must be indexed and eligible to show with a snippet in Google Search. To limit what is shown, Google points to nosnippet, data-nosnippet, max-snippet and noindex.

How do you get your brand mentioned once you know why?

Fix the cause your panel shows, starting with the sources: get your brand named, accurately and with disclosure, on the pages the engines already cite for your questions. Then make the entity unambiguous, open crawler access if it was closed, and publish pages that answer the open questions directly.

1. Sort the cited domains from your panel by how often they appear. For each one, find the realistic way in: a directory listing, a factual pitch to the author of a roundup, a post under your real name on the platform, a public repository if the directory is built from them. Disclose any relationship and any free licence.
2. Write one plain description of what you sell, who makes it, for whom and at what price, and use the same name and description everywhere you control: site, profiles, listings, structured data. If a similar name exists, say once that you are not affiliated with it.
3. Make your own sites corroborate each other: a page that states who publishes the product, with the legal entity and a contact, and links in both directions between the company site and the product site.
4. For questions answered without any brand, publish a page that answers that exact question in its first lines, with dated facts. One useful page per real question, not variants of the same page.
5. Re-run the frozen panel after four to six weeks and keep a dated log of what you changed, so that a movement can be traced to a cause.

• Street Cred: turns the cited sources where you are missing into a scored list of listings and pitches, with outreach drafts that say who you are and a follow-up tracker.

Does llms.txt, schema markup or a paid mention package help?

Not on their own. Google states that no machine-readable file, AI text file or special schema.org markup is needed to appear in AI Overviews or AI Mode, and Google's spam policies list buying links, including through paid posts, as link spam. None of them puts your name on the pages the engines cite.

llms.txt is a proposed plain-text summary of a site. It does no harm, but none of the crawler pages cited here says that their crawlers read it to choose what to cite. Publish one only if someone will keep it accurate.

Structured data can state explicitly that your name, your organisation and your profiles are the same entity, which helps when the name collides with others. It does not create mentions where no source names you.

Paid placements: Google's spam policies list exchanging money for links or for posts that contain links, and exchanging goods or services for links, as link spam. Hidden instructions for AI and mass-produced "best X" pages can fall under its policies on hidden text and scaled content abuse.

• Schema Sheriff: checks your JSON-LD against what the page shows, audits llms.txt, and answers first whether llms.txt is worth having.