Start free trial Dashboard →

Free WordPress plugin

Which AI crawlers really visit your WordPress site?

Anyone can call themselves GPTBot. This free plugin checks every claimed AI crawler visit against the addresses OpenAI, Anthropic, Perplexity, Google and Microsoft publish, and counts only the real ones.

AI crawler log Today

  • GPTBot OpenAI · address in OpenAI's list Verified
  • GPTBot Claimed · address not in OpenAI's list Spoofed
  • ClaudeBot Anthropic · behind a proxy, not checkable Unverified

79 verified fetches this month. 1 spoofed and 1 unchecked visit not counted.

Illustrative, as the plugin shows it on a test site.

Verification

Every claimed visit, checked against the operator's list.

Google tells site owners to check Googlebot by its address. OpenAI, Anthropic, Perplexity and Microsoft publish their crawlers' addresses too, so the plugin checks each claimed visit against them. A visit is called Spoofed only when there is no doubt.

A visit says

User-agent: GPTBot

Its address, checked against

openai.com/gptbot.json

Verified In the list. Counted.

Spoofed Not in it. Left out.

  • Verified

    The address is in the operator's published list. Only these visits are counted.

  • Spoofed

    It names a crawler, but the address is not in that operator's list. Left out of every count.

  • Unverified

    A proxy hid the real address, or a list could not be refreshed for a week. Never called spoofed.

  • Off until you say

    Checking means downloading those lists once a day, so it starts only when you turn it on, in one click.

Coverage

The pages AI has never read. Often the most useful list.

For each crawler you see whether robots.txt lets it in and the exact line that decides, which pages it fetched and when. Then the pages no verified AI crawler has fetched at all.

Pages AI crawlers fetched

Verified fetches, last 30 days

  • /blog/ai-search-guide 24 fetches
  • /services 9 fetches
  • /pricing Never fetched
  • /case-studies/acme Never fetched
Illustrative pages and counts.

Visits from AI answers

Visits from ChatGPT and Perplexity, counted per page.

Arrivals from ChatGPT, Perplexity, Gemini, Copilot, Claude and others, counted in the browser, so cached pages are included and a reload is not counted twice. A visit from Google AI Overviews looks like any other Google search visit, so it cannot be told apart.

Visits from AI answers, last 30 days

11

  • ChatGPT 7
  • Perplexity 4
/blog/ai-search-guide 6 from ChatGPT
Illustrative, as on a test site.

The crawlers

Every AI crawler, by name. And what blocking it costs.

Fifteen crawlers from OpenAI, Anthropic, Perplexity, Google, Microsoft and the training sets. Blocking a training crawler keeps you out of future models; blocking a search or user crawler takes you out of the answers people get today.

  • OAI-SearchBot

    OpenAI. Builds the index ChatGPT search answers from. Can send you visitors.

  • ChatGPT-User

    OpenAI. Opens a page when a ChatGPT user asks about it. Can send you visitors.

  • GPTBot

    OpenAI. Collects content to train OpenAI's models. Sends no visitors.

  • Claude-SearchBot

    Anthropic. Indexes pages for Claude's web search. Can send you visitors.

  • Claude-User

    Anthropic. Opens a page when a Claude user asks about it. Can send you visitors.

  • ClaudeBot

    Anthropic. Collects content to train Anthropic's models. Sends no visitors.

  • PerplexityBot

    Perplexity. Indexes pages Perplexity cites in its answers. Can send you visitors.

  • Perplexity-User

    Perplexity. Opens a page when a Perplexity user asks about it. Can send you visitors.

  • Googlebot

    Google. Search, including AI Overviews and AI Mode. Listed, never added to the AI totals.

  • Bingbot

    Microsoft. Bing, which Copilot and other assistants search through. Listed, never added to the AI totals.

  • Training sets

    Google-Extended, Applebot-Extended, Meta, Common Crawl and ByteDance. Training only.

Free fixes

Fixes you preview first. And can undo.

  • Structured data

    JSON-LD built only from what WordPress already knows about the page. Added only where none is output.

  • robots.txt

    Lets AI search crawlers in where robots.txt blocks them. Training crawlers are left alone.

  • llms.txt

    Generated for you. Offered as a convenience: no major AI engine has confirmed it reads one.

  • Headings

    A heading structure without skipped levels, proposed for you to review. Never applied without you.

Privacy

Nothing about people. Nothing sent to us.

  • No IP addresses

    Per crawler request: the time, the crawler, whether its address matched, the page and the status. Never the address.

  • No account

    Nothing to sign up for. Everything is worked out on your site and stored in your own database.

  • Nothing sent to us

    Its only requests are the operators' published lists, once a day, after you turn checking on.

  • Honest counts

    A page cache can answer without running WordPress. The plugin detects caches and says its counts are a floor.

Install

Running in a minute. Free, with no account.

WordPress 6.5 or later, PHP 8.0 or later. Works alongside Yoast, Rank Math, SEOPress and All in One SEO.

  1. Install the plugin

    In your dashboard, go to Plugins, Add New Plugin, search for AN10 AI Crawler Analytics, then install and activate it. Or download it from WordPress.org.

  2. Turn on verification

    One click on the plugin's screen. It downloads each operator's published list once a day. Turn it off in Settings and the lists are deleted.

  3. Read your crawler log

    Verified fetches by crawler and page, the pages none have fetched, and visits from AI answers, counted from the moment it is active.

Open source

The code is public. GPL, on GitHub.

Every change runs through its tests and WordPress.org's Plugin Check before it ships. Issues and pull requests are welcome.

Questions, answered.

Yes, all of it. There is no account, no paid version inside the plugin and no upgrade screen. It is open source under the GPL.
OpenAI, Anthropic, Perplexity, Google and Microsoft each publish the addresses their crawlers use. When a request says it is GPTBot, the plugin checks its address against OpenAI's list. If the address is not on it, the visit is marked Spoofed and left out of the counts. The address itself is never stored.
That is your call, and it is a different call from the others. GPTBot collects content for training. Blocking it does not remove you from ChatGPT's answers today; blocking OAI-SearchBot or ChatGPT-User does. Many sites refuse training crawlers and allow search and user crawlers.
All three are OpenAI's. OAI-SearchBot builds the index ChatGPT search answers from. ChatGPT-User opens a page when someone asks ChatGPT about it. GPTBot collects content to train OpenAI's models. Anthropic and Perplexity run the same three kinds.
Checking a visit is a lookup in lists already stored on your site. The lists are downloaded by a scheduled task once a day, never while a page is loading. Visits from AI answers are counted by a small script after the page has loaded.
Cloudflare is recognised from its published ranges. When a proxy hides a visitor's real address, the visit is marked Unverified rather than Spoofed, so a real crawler is never accused. A page cache that answers without running WordPress hides those visits from any plugin, so the screen says its counts are a floor.
No. Crawling is not citing, and finding out whether ChatGPT and Perplexity name you when a buyer asks means asking the engines, which the plugin never does. The free AI visibility audit on agentnumber10.com does that.
Yes. Your SEO plugin keeps its titles, meta descriptions, canonicals and sitemaps. This plugin never writes them, and adds structured data only where none is output.

See what the agent
would fix.

Try it on your own sites for 14 days. No card needed.