Guide · Directories and stores

Can the GPT Store and AI agent directories get you recommended by AI?

Updated 2026-11-09 · kagentx

Short answer

No guarantee. Nobody official says AI answers cite directory pages, and we have no evidence of actual citations. What we can confirm are the preconditions: on 2026-11-09 we tested three kinds of directory. A GPT Store page without logging in returns an app shell, with no GPT name or description in the raw HTML; Poe bot pages put the title and description in the HTML, but robots.txt blocks ClaudeBot and PerplexityBot from the whole site; kagentx's product pages put the name and description in the HTML and allow all three crawlers. Check whether the page can be read, then submit, and keep the description consistent with your site.

Will AI cite directory pages?

Nobody official has said so and we have no evidence of actual citations, so we cannot answer "yes". What we can check is the precondition: whether a crawler can read the directory page. On 2026-11-09 we fetched pages and robots.txt from several kinds of directory with curl (ordinary Mozilla/5.0 identity; what real crawlers see may differ):

DirectoryPage without logging inName and description in the raw HTML?GPTBotClaudeBotPerplexityBot
GPT Store (chatgpt.com/gpts and /g/…)200No: the <title> is just "ChatGPT", there is no meta description or canonical, and the GPT's name cannot be foundAllowAllowDisallow(点名整站禁止)
Poe bot page (poe.com/GPT-4o)200(首页有一次 403 验证页)Yes: <title> and meta description are both in the raw HTMLAllow(只禁 /p/*)Disallow(* 组整站禁止)Disallow(* 组整站禁止)
kagentx product directory (kagentx.com/p/…)200Yes: title, meta description and canonical; product pages also carry SoftwareApplication JSON-LDAllowAllowAllow

How to read it: Allow and Disallow are inferred from the robots.txt text, not from real crawler behaviour. We did not verify that the two GPT addresses are live GPTs; all we can confirm is that without logging in you get the same app shell, and what real crawlers see may differ, which is unverified. For robots.txt syntax see How do you write robots.txt for AI crawlers?; for the effect of script rendering see Do AI crawlers render JavaScript?.

This shows two things: directories differ a lot, and not every directory page can be read; and putting a page somewhere outside your own site does not mean AI has seen it. So check first, then submit.

Which kinds of directory should you submit to?

You do not need to do everything at once. Use the three kinds below, and in each pick the ones whose pages can be read first:

CategoryExampleBest forCheck before submitting
Stores built into a platformGPT Store, PoeYou already built a GPT or bot on that platformWhether the page title and description are in the raw HTML, and what robots.txt says
AI tool or agent directorieskagentx.com/agents/Indie developers building AI tools or agentsWhether each entry has its own URL, shows name and description in the HTML, and links back to your site
Scorecards and ranking sitespagent.kagentx.comYour product belongs to a category it testsWhether the inclusion criteria and scoring method are clear, and whether each result can be checked

Scorecards, ranking articles and review sites work differently: see How do I get into “best X tools” list articles? and Can G2, Capterra and Trustpilot reviews make AI recommend you more often?; for AI tool products in general see How do AI tool products get recommended by ChatGPT?. Pagent (a scorecard for personal AI assistants you can message) is a scorecard we built, in English, used here only as an example.

  1. 1. Check whether the page can be read

    Run these lines on other people's entry pages in the directory:

    # Can a crawler read the name and description without logging in?
    curl -s https://example-directory.com/entry/your-product | grep -io '<title>[^<]*</title>'
    curl -s https://example-directory.com/entry/your-product | grep -io '<meta name="description"[^>]*>'
    # What does the directory's robots.txt say about the AI crawlers?
    curl -s https://example-directory.com/robots.txt | grep -i -A3 -E "gptbot|claudebot|perplexitybot|user-agent: \*"
  2. 2. The entry must link to your site

    Prefer an ordinary link in the body text. That way, when AI sees the directory entry, it can find your original source.

  3. 3. Submit only to categories that fit your product

    If you pick the wrong category your entry shows up in unrelated lists, which is no use to buyers.

  4. 4. Keep a table of where you submitted

    Write the directory name, entry URL, submission date and the description used then. You will need it later to check consistency and whether you were indexed.

How do you keep directory descriptions consistent with your own site?

Write one shared copy and use it on your site and every directory entry. What should be identical is the name, one-line description, category, URL, pricing wording and update date:

Name:        Example Slides
One line:    Turns a PDF into an editable slide deck, for students and consultants.
Category:    Presentation tools
URL:         https://example.com/
Pricing:     Plans listed at https://example.com/pricing/ (checked 2026-11)
Updated:     2026-11-09
Same words used on: homepage, About page, every directory entry

Three rules:

  1. 1. Your site is the source, directories copy it

    Use the same sentence on your home page and in How to write an About page so AI gets your company and product right. Copy that sentence into each directory entry instead of rewriting it for each one.

  2. 2. One pricing wording, linked to the pricing page

    Do not spell out pricing details in the directory; write "see the pricing page on our site" and give the URL. When prices change you edit one place.

  3. 3. After every product change, recheck the directories

    Using the table above, go through each directory and see whether the description is still right. An outdated description will be taken by AI as what you say now.

What is the situation with kagentx's own product directory?

This is a directory we run ourselves. Below are only facts found on 2026-11-09; we do not infer whether AI cites it:

ItemResult
Product entries in the library (web/data/products.json)1,155
Of those, product pages with the full analysis (verdict, value chain, competition, risks) that may be indexed: how many the live /agents/ list and sitemap.xml each hold141
The other product pagesThey open, but carry noindex and stay out of the sitemap (the rule in the build script web/scripts/build-agents-seo.mjs, read on 2026-11-09)
Product page /p/lindy (has the full analysis, indexable)200 without logging in; title, description and canonical are in the server-rendered HTML, no robots meta; also has SoftwareApplication structured data
Product page /p/research-stammer (no full analysis)200 without logging in; title, description and canonical are in the server-rendered HTML, but it carries noindex, follow (robots meta) and is not indexed (re-checked 2026-11-10)
robots.txtAllow: / for all crawlers; only /admin.html, /candidates.html and /api/ are disallowed
Google Search ConsoleIn the last 28 days a few /p/research-* product pages had impressions (operational data; no figures given here)

So what we can say is: kagentx's product pages have impressions in Google, and the pages can be read without logging in. Whether ChatGPT, Perplexity or Claude have cited them we have not checked, so it is unverified. kagentx.com/agents/ is the list page.

How do you test after submitting?

  1. 1. Within a week: is the directory page indexed?

    Search a search engine for "your product name + directory name" and see whether the entry appears.

  2. 2. Write down 5 questions buyers would ask

    Leave your brand name out, phrase them the way buyers do, and ask each one 3 times with web search on.

  3. 3. Record the answers and sources

    Whether you are mentioned and whether the source list contains a directory page. For the method and a table, see How do you measure AI visibility? Four metrics and a table you can follow.

  4. 4. Look at your own site's referral data

    In GA4 or your server logs, look for visits coming from directory pages.

For apps built from an MCP server and skills in ChatGPT (called plugins in the docs now), how they are found and submitted for review, see How do users and AI find apps inside ChatGPT (Apps SDK, the plugin directory)?.

FAQ

Can GPT Store pages be read by search engines and AI?

On 2026-11-09 we fetched chatgpt.com/gpts and two GPT addresses with curl: the status was 200, but the title in the raw HTML is just "ChatGPT" and the GPT's name and description cannot be found. robots.txt allows GPTBot and ClaudeBot and blocks PerplexityBot from the whole site. What real crawlers see is unverified.

Do AI answers cite AI agent directory pages?

Nobody official says so and we have no evidence of actual citations, so this is unverified. What we can confirm are the preconditions: the directory page must be readable without logging in, the name and description must be in the raw HTML, and robots.txt must not block the crawlers.

Which kind of directory should I submit to first?

Start with those whose pages can be read: use the curl lines above to see whether the title and description are in the raw HTML, then read robots.txt. A directory that cannot be read cannot be seen by crawlers, however good the page is.

Does a directory description have to be identical to my site's?

Keep the name, one-line description, category, URL and pricing wording consistent; the rest can differ. When they disagree, AI sees several contradictory claims and does not know which to trust.

How do I know whether it worked after submitting?

First see whether the directory page shows up in search results, then ask AI the buyers' questions and check whether you appear in the answer and sources; for the method see "How to measure AI visibility". One result is not a conclusion: ask the same question several times.

Sources

kagentx

Do AI answers
mention you?

The evaluation checks whether crawlers can get in and whether the text is served directly, then asks ChatGPT, Perplexity and Claude (with web search on) your buyers' questions and sees whether you are mentioned. The Basic plan is $1, and results usually take 2–3 minutes.