Why doesn’t ChatGPT recommend your business?
Your customers are beginning to ask an assistant to shortlist vendors for them. That assistant visits your site, tries to work out what you sell, what it costs and how to start — and if it can’t, it moves on to whoever it can. You never see the enquiry, because there was never an enquiry.
Pointer is a 20-second check that tells you, in plain English, what stops it. Enter your domain and read the report. The report is free.
Why you may already have a good score somewhere else
There are several AI-readiness scanners. They are useful and we recommend running them. But every one we have tested asks the same question: do the right files exist? Is there a robots.txt, a sitemap, an llms.txt, some structured data.
None of them call the things those files promise.
A site can publish a flawless description of an API that was never built, and score a hundred everywhere, while an agent trying to use it is told a purchase succeeded when nothing happened. Files existing is not the same as the site working.
We scored our own site 39 out of 100
On the day we built this, a well-known scanner rated 3dogs.ai 90 out of 100, “AI-Ready”. Pointer scored the same site, in the same minute, 39 — and it was right. We were advertising a checkout API that did not exist, pointing agents at documentation that returned our homepage, and answering “success” to every dead link on the domain.
None of that showed up on the discovery-layer score, because none of it is a missing file. All of it would have broken a real agent trying to transact with us. We are publishing our own number rather than a flattering one, and you can re-run it yourself at any time.
What Pointer actually checks
Eighteen live requests in six categories. Two of them — whether you can be found, and whether your content can be read — are the ones other tools cover. The other four are ours:
- Truthfulness. We call every endpoint your API description declares, and follow every link your agent card advertises. A promise that does not answer is a finding, not a footnote.
- Authority. Operations that change something should say who is allowed to invoke them. We check that the permission is declared, not assumed.
- Failure honesty. We request a URL that cannot exist. If your site answers “success”, an agent has no way to tell your real pages from your dead ones.
- Commerce. Can an agent find out what you charge, and what to do next? If not, it cannot put you on a shortlist.
Every finding on the report shows the exact request that produced it. A document describing an API is evidence of a document.
Then we fix it — and prove it
Most findings are not a file you can paste. A soft-404 lives in your router. An API you advertise but never built is a decision about what your company actually offers, and no template can make that decision for you. That is where the other tools stop, and it is where the work starts.
Included with Corporate. We publish the files we can generate, make the hosting changes with your team or hand them over specified and scoped, walk you through the judgement calls, then re-run this exact scan and put the before and after in writing. We keep monitoring it and re-scan on a schedule so it doesn’t quietly break again. The report you get for free; what you are buying is the second scan coming back clean — and Corporate includes a guarantee that we will improve your agent-readiness score, not that we will make it 100.
The work is carried out by the Malamute Harness — our build team of independent foundational models with a human in the loop.