When a model like ChatGPT decides whether to mention your company, it isn't ranking you the way a search engine does. It's assembling an answer from what it can find, read, and trust about you — and most sites give it surprisingly little to work with.
A ChatGPT-readiness scan is a way of seeing your site the way the model does. Not a vague "AI score," but a concrete inventory: which signals exist, which are missing, and which gaps a competitor's page is currently filling on your behalf.
01What the scan reads
The scan crawls your site the way an AI system's retrieval layer would: the homepage, your highest-traffic pages, and the machine-facing files most people never look at. For each page it asks three questions:
- Can a model find this? Is the page crawlable, linked, and present in your sitemap — or buried behind navigation only a human would click?
- Can a model read this? Is the substance in real text and structured data, or locked inside images, scripts, and vague headings?
- Would a model trust this? Does the page say plainly who you are, who you serve, and how you compare — with enough specificity to quote?
The single most common finding across scans: the model can read your homepage — and not much else. Pricing, comparisons, and "who we serve" detail usually live in a deck, a sales call, or someone's head.
02The signal groups, in plain terms
Scan findings cluster into four groups. You don't need to memorize them — but knowing the groups makes a 47-finding report feel a lot less alarming.
- Technical signals. llms.txt, robots rules, sitemap coverage, canonical tags. Cheap to fix, foundational to everything else.
- Structured data. Organization, FAQ, Product, and Article schema — the fields models lean on when they summarize you.
- Business context. Plain-text answers to "what is this company, who is it for, what does it cost, how is it different."
- Content coverage. The questions your market asks that your site doesn't answer — your content gaps.
03What "missing" looks like
Most flagged gaps aren't broken pages — they're absent files. Take llms.txt: a short, plain-text map that tells a model what your site contains and where the substance lives. A useful one is modest:
Ten minutes of writing. Yet in scans of small-business sites, it's missing far more often than it's present — along with FAQ schema and any page a model could quote when asked "Acme vs. its alternatives."
Curious which of these signals your site is missing? The free scan checks all four groups and returns one live snapshot — no signup.
Scan my website04Fix order: what moves first
Not all gaps deserve equal urgency. This is the order we'd work through a typical scan result — fastest readiness movement first:
| Priority | Fix | Effort | Why first |
|---|---|---|---|
| Now | llms.txt + sitemap coverage | Hours | Unblocks everything downstream — a model can't weigh pages it never finds. |
| Now | Organization + FAQ schema | Hours | Gives models quotable, structured facts instead of guesses. |
| This week | Plain-text business context | Days | "Who we serve" and "what it costs" in real text, on real pages. |
| This week | Title + metadata rewrites | Days | Vague titles read as vague companies. Specific ones get summarized correctly. |
| Ongoing | Content gaps → growth assets | Weekly | Comparison pages, FAQs, and topical pages — the slow compounding layer. |
Resist the urge to start with content. Technical signals are an afternoon; content is a quarter. Ship the afternoon first so the quarter's work lands on a readable site.
05From fixes to a Growth Plan
One-off fixes raise the floor. What raises visibility over time is treating the rest of the scan — the content gaps — as a queue of work. In AI Friendly, that queue becomes a 30-day Growth Plan: each gap mapped to a growth asset, drafted as a reviewable brief, and delivered as a publish-ready payload you approve before anything ships.
You don't have to use our planner to apply the idea. The discipline is the same anywhere: rank gaps by how often your market asks the question, produce one asset at a time, and review everything before it goes live.
06Track whether it worked
Readiness work without tracking is decoration. After the first wave of fixes, watch three things — none of which require a dashboard full of charts:
- Re-scan delta. Did the missing-signal count actually drop after deploy?
- Prompt coverage. When you ask assistants the questions your buyers ask, do you appear — and is what they say accurate?
- Assisted traffic. Referrals and brand searches that follow AI mentions, trended monthly, not daily.
Visibility movement is gradual and depends on your market and competition — anyone promising a guaranteed citation is selling something else. What you can guarantee is the inputs: signals present, context readable, gaps closing week over week. That's the whole game.