Invisible in AI Answers? Test Branded Prompts, Then Category
An "invisible" result means an AI assistant answered one of your target prompts and your brand did not appear anywhere: not cited as a source, not named in the text, not on any shortlist. When that happens across every assistant and most prompts, the cause is rarely one missing tag.
This guide shows you how to tell whether the assistants do not know your business exists, know it but cannot describe it, or know it but never think of it for your category, and what to fix first in each case so the next round of tests shows movement.
What an "invisible" result actually tells you
The AI Visibility Tracker sends your target prompts to several assistants and records one of four outcomes for your brand in each answer. They are not a ranking; they describe different relationships between the assistant and your business.
| Status | What happened in the answer | What it suggests |
|---|---|---|
| Cited | One of your pages is listed or linked as a source. | The assistant found and used your content for this prompt. |
| Mentioned | Your brand is named in the text, with or without a link. | The assistant knows you exist and connects you with the topic. |
| Recommended | Your brand is suggested as an option the reader should consider. | You are treated as a credible choice in your category. |
| Invisible | None of the above. | Either the assistant has no usable information about you, or other businesses fit the prompt better in its sources. |
One invisible answer means very little. AI answers are generated fresh each time, and the pages an assistant retrieves can change from one run to the next. The finding that matters is a pattern: invisible on most prompts, across most assistants, over several weeks of runs. That pattern almost always comes from one of three gaps, and the order in which you fix them matters because each depends on the one before.
- A discovery gap. The assistants, and the search indexes they draw on, have little or nothing about your business. Nothing downstream can work until this is fixed.
- A description gap. The assistants know the name but cannot say clearly what you do, where, or for whom, so they cannot match you to a question.
- A prominence gap. The assistants know exactly what you do, but the sources they read for category questions talk about other businesses, so your competitors fill the answer.
Branded, category and comparison prompts
The fastest way to find which gap you have is to sort your prompts by what they ask. Most businesses start tracking only category prompts, the questions a new customer asks before they have heard of anyone, and then conclude the assistants ignore them. Adding branded prompts to the set turns a vague result into a diagnosis.
| Prompt type | Example for a fictional firm | If you are invisible here |
|---|---|---|
| Branded | "What does Fernbrook Joinery do?" | A discovery or description gap. The assistant cannot find, or cannot trust, information about you even when handed your name. |
| Branded plus attribute | "Does Fernbrook Joinery make fitted wardrobes in Derby?" | A description gap. The assistant knows the name but not the specifics that would let it match you to a need. |
| Category | "Who makes bespoke fitted wardrobes near Derby?" | A prominence gap, if branded prompts work. The sources for this question do not include you. |
| Comparison | "Fernbrook Joinery or Oakhill Interiors for alcove units?" | Usually a description gap: the assistant cannot say enough about you to compare. |
Read the results as a decision tree. Invisible on branded prompts means stop working on anything else and fix discovery. Visible on branded prompts but vague or wrong means fix the description. Clear on branded prompts but invisible on category prompts means the work is off your site, in the pages the assistants read when they research your category.
Where each assistant looks for information
Assistants answer from two places: what the model learned in training, and what it retrieves from the web at the moment you ask. For current, local and commercial questions, retrieval usually does most of the work, and retrieval depends on an index. The companies have published some details about their crawlers and sources, but none of them publishes a full list of how sources are chosen, so treat the table below as the parts that are documented, not a complete recipe.
| Assistant | What it can draw on (as documented) | What to check |
|---|---|---|
| ChatGPT | OpenAI's own crawlers, including OAI-SearchBot for search and GPTBot
for training, plus third-party search providers. | robots.txt and firewall rules for OpenAI's user agents; Bing indexing is a sensible extra check. |
| Gemini and Google AI features | Google's search index, built by Googlebot.
The Google-Extended token governs some Gemini uses but not Search. | Search Console indexing for your key pages. |
| Copilot | Microsoft's Bing index. | Bing Webmaster Tools: is the site verified and are the pages indexed? |
| Perplexity | PerplexityBot for its index, and Perplexity-User
when fetching a page for a user. | robots.txt, firewall and CDN bot rules. |
| Claude | Anthropic documents ClaudeBot, Claude-SearchBot and
Claude-User, and web search when it is enabled. | robots.txt and firewall rules for Anthropic's user agents. |
| Brave Leo | Brave's own search index. | Whether your pages appear in Brave Search results. |
The practical point is that "AI visibility" rests on ordinary indexes. If Bing has not indexed your site, Copilot has little to retrieve. If Google has not indexed your service pages, Gemini and AI Overviews cannot draw on them. If your firewall sends a challenge page to Perplexity's crawler, Perplexity sees the challenge, not your content. A quick way to check the search layer is a plain search for your exact business name in Google, Bing and Brave Search: if your own site is not on the first page for your own name, the assistants that rely on those indexes will struggle too.
The fix, step by step
Step 1
Split your prompts into branded, category and comparison
Rewrite your tracked prompt set so it has three groups: about five branded prompts that name your business, ten to twenty category prompts written the way a customer who has never heard of you would ask, and a few comparison prompts that pair you with a known competitor. Tag each prompt with its group in the tracker or your spreadsheet, so every later result can be read by group rather than as one overall score.
Step 2
Ask each assistant about your brand by name
Run the branded group across every assistant and read each answer in full. Note whether the assistant recognises the business, whether it describes the right one, and which facts it gets right or wrong: what you do, location, services, founding year, owner. If most assistants say they have no information, or describe a different company with a similar name, you have a discovery gap and steps three and four come before anything else.
Step 3
Check the index each assistant draws on
Search your exact business name in Google, Bing and Brave Search, and check Search Console
and Bing Webmaster Tools for indexing of your home, about, contact and main service pages. Then read
your robots.txt, firewall and CDN bot settings for rules that block OAI-SearchBot,
PerplexityBot, Claude-SearchBot, Bingbot or Googlebot.
Confirm in your server logs that those crawlers receive a normal page with a 200 status, not a challenge
or error page.
Step 4
Correct what assistants say when they do know you
Where branded answers are vague or wrong, make one page on your site state the
core facts plainly in the opening lines: the full business name, what you do, who for, where, and since
when. Match those facts exactly on your Google Business Profile, social profiles and main directory listings,
add Organization or LocalBusiness markup with sameAs links to
those profiles, and remove old pages that still show former names, addresses or services.
Step 5
List the pages that win your category prompts
For each category prompt where you are invisible, record every source the assistants cite and every business they name. Sort the sources into types: best-of lists, directories, review sites, forum threads, news articles and competitors' own pages. Count how often each source and each competitor appears across prompts and assistants. The sources that keep recurring are the pages that shape your category, and they tell you where your business needs to be mentioned.
Step 6
Get your business into the sources those answers use
Work down the recurring sources from step five, starting with those you can influence directly: claim or complete directory and review-site listings, ask real customers to leave reviews on the platforms that appear, and contact the authors of best-of lists with accurate information about what you offer. Where the winning source is a competitor's page answering the exact question, publish a better page of your own that answers it directly and specifically.
Step 7
Re-run the same prompts on a fixed schedule
Keep the prompt wording, assistants and settings the same, and run the set on a fixed day each week or fortnight. Read the results by group: branded prompts should improve first, usually within weeks of indexing fixes, while category prompts move more slowly as third-party sources are updated and recrawled. Record each change you make with its date, so a later movement in the tracker can be matched against the work that probably caused it.
Reading the pattern across assistants
The tracker records results for each assistant separately, and differences between them are often the most useful clue. Because the assistants draw on different indexes, a gap that appears in only some of them points at the index they share.
- Visible in Gemini, invisible in Copilot. Suspect Bing. Check that the site is verified in Bing Webmaster Tools, that the sitemap has been submitted there, and that Bing has indexed your key pages. Sites that only ever set up Google Search Console often have thin Bing coverage.
- Visible in Copilot and Gemini, invisible in Perplexity and ChatGPT. Suspect a crawler block. Those two run their own crawlers, and a firewall rule, bot-fight setting or robots.txt line that singles out AI user agents will hide you from them while leaving search engines untouched.
- Invisible everywhere on branded prompts, but your site ranks for your name. The assistants may be matching your name to something else, or the pages that rank say too little about what you do. Look at step four.
- Named in the text but never cited. The assistant knows you from other sources, such as directories or reviews, but did not use your own site. That is a description or content problem on your pages, not an invisibility problem.
- Visible when the assistant searches, invisible when it does not. Some assistants answer some prompts from training alone. Newer businesses are unlikely to be in training data, so prompts that trigger a web search are the realistic target.
Cases that need a different approach
A very new business
If your site is a few months old, invisibility is expected. There is little for any index to hold and few other sites mention you. Focus on steps three and four so that branded prompts work, and treat category visibility as a longer project built on reviews, listings and coverage that take time to accumulate.
A name shared with something else
If your business name is also a common word, a place, or another company's name, assistants may answer
confidently about the wrong thing. Use your full name consistently, including any descriptor such as
"Fernbrook Joinery" rather than "Fernbrook", make your about page state what you are not as well as what
you are where confusion is likely, and add alternateName only for names you genuinely use.
A local service business
Local questions lean heavily on maps listings and review platforms. A complete Google Business Profile, a Bing Places listing, and consistent name, address and phone details across directories often matter more than anything on the website itself.
A recent rebrand
Assistants may still know you only by the old name. Keep a clear line on your about page such as "formerly known as", redirect old URLs, and update the old name on profiles you control. Track both names as branded prompts until the new one is recognised.
A niche business-to-business supplier
Category prompts in narrow industries often return only a handful of well-known names, drawn from trade directories, association member lists and industry publications. Those are the sources to work on, rather than general consumer review sites.
What will not fix invisibility
Several popular shortcuts are either unproven or likely to make things worse. None of the assistant providers has published a way to request inclusion in answers, and nobody outside those companies can guarantee a mention.
- Publishing large volumes of generated articles. More pages do not create a reason for anyone to mention you, and low-value content can hurt how search engines treat the site as a whole.
- Fake reviews or planted forum posts. Review platforms and forums remove them, they can breach consumer protection law in the UK, and they give assistants misleading information about you.
- Relying on an llms.txt file alone. It can help some tools find your key pages, but it does not put your business into search indexes or third-party sources, and support for it varies.
- Serving different content to AI crawlers. Showing crawlers something visitors do not see is cloaking, which search engines treat as spam.
- Judging by a single answer. An answer that includes you once, or leaves you out once, proves nothing. Only repeated runs of the same prompts show a trend.
Stopping it coming back
Invisibility often returns quietly after a change that nobody linked to AI visibility. A few habits catch it early:
- Keep the branded prompts in your tracked set permanently. A sudden drop on branded prompts is an early warning that something technical has changed.
- After any change to your firewall, CDN, hosting or security plugin, recheck that the crawlers listed above still receive a normal page.
- When you change your name, address, phone or services, update your site, your Google Business Profile, Bing Places and main directories in the same week.
- Review the recurring sources from step five every quarter. New best-of lists and comparison pages appear, and old ones stop being cited.
- Keep a simple change log alongside the tracker results, so you can see which work coincided with which change in visibility.
1. How long does it take to stop being invisible in AI answers?
It depends on which gap you have. If the cause is a crawler block or missing indexing, branded prompts can start working within a few weeks of the fix, once the crawlers revisit and the indexes update. Category visibility takes longer because it depends on other websites mentioning you, and those pages are written, updated and recrawled on their own schedule. Plan in months rather than weeks for category prompts, and judge progress by the trend across repeated runs.
2. Why does ChatGPT know my competitor but not me?
Usually because your competitor appears in more of the pages ChatGPT retrieves for your category: directories, review sites, comparison articles, forum threads and news coverage. It can also be that your competitor has been trading longer, so it may appear in training data too. Check the sources cited when ChatGPT names the competitor, then work on appearing in the same kinds of sources. Also rule out a block on OpenAI's crawlers, which would hide your own pages.
3. Is it worth tracking branded prompts if customers do not know my name?
Yes, because branded prompts are diagnostic rather than commercial. If an assistant cannot describe your business when given its name, it has no chance of recommending you for a general question. Branded prompts also react quickly to technical problems, so a sudden change there warns you that something has broken. Keep a small branded group alongside your category prompts and read the two groups separately rather than as one blended score.
4. Does my site need to rank in Google to appear in AI answers?
Not in every assistant, but search indexes matter a great deal. Gemini and Google's AI features draw on Google's index, Copilot on Bing's, and Brave Leo on Brave's. Pages that are not indexed cannot be retrieved by assistants that rely on those indexes. Ranking at the top is not a strict requirement, since assistants often cite pages from beyond the first few results, but being indexed and readable is the starting point.
5. Will blocking GPTBot make me invisible in ChatGPT?
Blocking GPTBot tells OpenAI not to use your content for training future models. OpenAI documents a separate crawler, OAI-SearchBot, for its search features, and a ChatGPT-User agent for fetches made on behalf of a user. If you want to appear in ChatGPT search answers, check that OAI-SearchBot is allowed in robots.txt and not blocked by your firewall. Decide on GPTBot separately, according to whether you are comfortable with training use.
6. Why do I appear in Perplexity but not in Gemini?
The two draw on different indexes. Perplexity runs its own crawler and index, while Gemini and Google's AI features rely on Google Search. If Perplexity cites you but Gemini does not, look at Google indexing first: check Search Console for your key pages, look for noindex tags or canonical tags pointing elsewhere, and confirm Googlebot receives the same page visitors see. Differences in how each assistant chooses sources also play a part.
7. Can I pay to be included in AI assistant answers?
Some assistants have begun testing advertising, but ads are labelled separately from the answer, and none of the providers offers a paid route into the organic answer text or its citations. Services that promise guaranteed mentions cannot control how the assistants choose sources. Your money is better spent on the things the assistants demonstrably read: an indexable, clear website, complete listings on the platforms they cite, genuine reviews and coverage in publications your customers trust.
The AI Visibility Tracker asks ChatGPT, Claude, Perplexity, Gemini, Copilot and Brave Leo your target prompts, records whether your brand is cited, mentioned, recommended or invisible, tracks how often over time and shows the gaps against competitors.
Open AI Visibility Fixes →