| Free tool

Is your website visible to AI?

People are asking ChatGPT and Claude for recommendations instead of searching. If those assistants cannot open your website, they cannot recommend your business — and most owners have no idea whether they can. Check yours below in about twelve seconds.

3AI crawlers tested
~12sto the answer
Freeno signup, no card
Fixableusually one setting

The check below asks for your page as GPTBot, ClaudeBot and PerplexityBot and reports what your server actually said to each one. It also runs the full website audit, so you get the speed, security and search findings at the same time.

| The reason this page exists

We found it on our own site, by accident.

On 25 September 2026 we tested this site the way the tool above does it — by asking for a page while pretending to be each AI crawler in turn. This is what came back:

403  GPTBot         ← ChatGPT’s crawler, refused
403  ClaudeBot      ← Claude’s crawler, refused
403  CCBot          ← Common Crawl, refused
200  PerplexityBot
200  Googlebot

Our robots.txt explicitly welcomed all of them. The block was a firewall setting we had never opened — switched on by default — and it had been quietly turning away two of the three biggest AI assistants. Claude could not read a single page of this site.

Nothing warned us. Google was fine. Visitors were fine. Analytics looked normal, because a refused crawler is not a visitor and never shows up in a report. The only way to find it was to ask as the crawler, which is precisely what almost nobody does.

It was one setting, and it took a minute to change. That is the part worth knowing: this is rarely a difficult problem. It is an invisible one.

| How it happens

Four ways a site blocks AI without anyone deciding to.

01

A firewall default

Cloudflare and several other providers ship with AI-crawler blocking available, and it is easy to switch on without realising it also covers the assistants people ask for recommendations. This is what had happened to us.

02

A security plugin

WordPress security plugins often bundle "bot protection" that treats unfamiliar crawlers as threats. GPTBot and ClaudeBot are unfamiliar to a plugin written before they existed.

03

A robots.txt somebody copied

Blocking AI crawlers was widely recommended in 2023 and 2024, and a lot of those lines were pasted into robots.txt files and forgotten. Worth re-reading yours now that the trade-off has changed.

04

Hosting-level rate limiting

Some shared hosts throttle anything that crawls quickly. The crawler sees 429, gives up, and comes back less often — or stops coming.

| If you are blocked

How to fix it, by cause.

Find which of the four it is before changing anything — unblocking in the wrong place leaves the real block in place and makes you think it is fixed.

  • Cloudflare: Security → Bots, and look for the AI scrapers and crawlers setting. Ours was on ai_bots_protection. Turning it off restored every crawler immediately. If you use the API rather than the dashboard, note that ai_training is a read-only field derived from that setting — writing to it directly returns a Bad Request.
  • A security plugin: find its bot or crawler blocklist and allow GPTBot, ClaudeBot, OAI-SearchBot, PerplexityBot and Google-Extended.
  • robots.txt: look for User-agent: GPTBot or ClaudeBot followed by Disallow: /. Remove those blocks, or keep them deliberately — but then do not also expect to appear in AI answers.
  • Rate limiting: usually a hosting support ticket rather than a setting you can reach.
This is a genuine choice, not an obvious win. Letting AI crawlers in means your content can be used to train models as well as to cite you. Some businesses would rather not, and that is a legitimate position. What is not legitimate is a robots.txt that says one thing while the server does another — whichever you choose, make the two agree.

| Beyond access

Being readable is the floor, not the goal.

Once assistants can open your pages, the question becomes whether they can use what they find.

01

Answer questions in plain sentences

An assistant lifts a sentence that answers the question directly. "Website maintenance in Pakistan costs PKR 9,000 to 35,000 a month" can be quoted. "Our pricing is tailored to your needs" cannot be quoted by anyone, about anything.

02

Mark up your facts

Organization or LocalBusiness markup for your business details, FAQPage for your questions and answers. This is the difference between an assistant finding you and an assistant quoting you with confidence.

03

Publish the numbers you are willing to publish

Prices, hours, response times, service areas. An assistant asked "who does X in Lahore and what does it cost" will name the business that answered, not the one that said "contact us for a quote".

04

Add an llms.txt

A plain text file at yoursite.com/llms.txt describing what you do and which pages matter. Nothing requires it. It takes an afternoon, and it means the summary an assistant reads is one you wrote.

05

Keep it true

Assistants cross-check. A claim on your site that contradicts your Google profile, your invoices or another page of your own is worse than no claim, because it makes everything near it less quotable.

| The words

SEO, AEO, GEO and LLMO, without the jargon.

  • SEO — search engine optimisation. Getting your page into a list of blue links.
  • AEO — answer engine optimisation. Getting your answer used when somebody asks a question rather than searching for one.
  • GEO — generative engine optimisation. Getting your business named inside an answer an AI writes from scratch.
  • LLMO — the same idea as GEO, different acronym. The industry has not settled on one yet, and you can safely ignore the distinction.

They are marketed as four disciplines. In practice they rest on three things: pages an assistant can fetch, facts it can lift without guessing, and claims that hold up when checked. Everything else is detail.

Not sure what your result means?

Send us the address on WhatsApp. We will run it ourselves and tell you plainly whether anything needs doing.

WhatsApp us