About this plugin
People increasingly ask an AI assistant to recommend a business instead of searching for one. Before an assistant can recommend you, it has to read your website.
Your AI visibility starts here. If ChatGPT, Claude, Perplexity or Google’s AI cannot reach your pages, nothing else you do for AI search matters, because there is nothing for them to read.
Plenty of sites block those assistants without the owner ever knowing. txt rules can all turn AI crawlers away while ordinary visitors see nothing wrong. If that is happening to you, no amount of good content will help, because nothing is being read.
That is not a guess. txt rule about AI at all. Nobody chose it and nobody was told.
io/
This plugin checks, from your own dashboard:
- Whether the agents that fetch your page when a customer asks (OAI-SearchBot, ChatGPT-User, Perplexity-User, Claude-SearchBot) are being served or refused
- Separately, whether the crawlers that collect pages to train on (GPTBot, ClaudeBot, PerplexityBot, Google-Extended) are being served or refused
- Whether your robots.txt tells any of them to stay out
- Whether WordPress’s own “Discourage search engines” setting is on
- Whether your front page carries a noindex tag telling engines not to list it
- Whether your front page has structured data for an assistant to lift facts from
- Whether you have an llms.txt file
It then gives you one plain sentence saying where you stand.
Honest about what it finds
A refused crawler is not always a deliberate block, so this plugin does not shout about one. It also requests your site as Googlebot as a control. If Google is being refused too, you have a broken bot rule rather than an AI policy, and the plugin says exactly that, because that problem is bigger and costs you traffic today.
It only counts 401, 403 and 429 as a deliberate refusal. A 400 is a malformed request and a 500 is your server erroring, and calling either a block would be misleading.
txt reader follows the real rules rather than guessing. A commented-out line is not a block. A rule written for your site as a whole does not apply to a crawler that has its own rules further down.
An Allow that cancels a Disallow is respected. Each of those, done carelessly, invents a block that is not there.
txt line behind it, so you can verify any claim it makes rather than taking it on trust.
Two different crawlers, two different consequences
Most tools in this space report one number for “AI crawlers” and leave you to
assume the worst. There are really two kinds, and only one of them decides
whether you can be recommended.
Some agents fetch your page at the moment a person asks an assistant about your
business. If those are refused, the assistant has nothing to read and cannot
describe or recommend you. That is worth fixing.
Others collect pages in bulk to train models on. Refusing those is a deliberate
choice a lot of sites now make, and on its own it does not stop anyone
recommending you.
This matters more than it sounds. Across 293 local business websites checked in
txt, and 14 of those 15 refused
only the training crawlers while deliberately allowing the search-time agents by
name. Reporting those 14 as “blocked from AI” would tell a business it is
invisible when it is not.
So this plugin tells you which kind you are refusing, and only raises an alarm
for the kind that costs you something.
Why there is no score
Other plugins in this space hand you a number out of 100 and a letter grade. This one does not, on purpose.
Every question it asks has a real answer. A crawler was served or it was refused. txt rule exists or it does not.
Averaging answers like those into a grade throws away the part you can actually act on, and it makes a short list of plain facts look like a measurement it is not.
txt line behind any claim. Every one of them is something you can check yourself.
What it cannot tell you
Requests come from your own server. If your site sits behind a CDN or firewall and your server bypasses it when calling its own address, a block at that edge will not appear here. Checking from outside your hosting is the only way to be sure.
A slow site gets a partial answer. The whole run is held to a wall clock budget so it cannot outlast PHP’s execution limit and leave you with a blank admin page. If your site is slow to answer requests from itself, the run stops early, the rows it never reached say “Could not test” instead of guessing, and it tells you it stopped.
txt only. There is no such crawler to refuse. Google reads your pages as Googlebot and then consults the Google-Extended rule to decide whether that content may feed Gemini and AI Overviews.
Any tool that reports an HTTP status for Google-Extended is reporting a request it invented.
Privacy
The plugin only requests your own site. It sends nothing to any third party, requires no account, no API key and no signup, and stores nothing beyond a one hour cache of your last result.