The AI Visibility Report, page by page.
The $299 report, shown on a demonstration store exactly the way it lands for you. It measures whether answer engines can reach you, read you, and cite you when a buyer asks. Then it turns every gap into a four week plan your developer can run.
SAMPLE · demo-outfitters.com, a demonstration store · 12 pages read · 16 named AI agents asked · 24 answers from a live engine
The front page of the report.
Six areas, each scored from something measured on the store’s live pages rather than from a rule of thumb. The last one decides whether an answer engine ever says your name, so the report prints it on its own and never averages it away against the areas that scored well.
Product pages lose their content without JavaScript
The catalog renders from script. Loaded a second time with JavaScript off, 8 of the store’s product pages keep their heading and lose the price, the sizes, and the description: exactly the parts an answer would quote.
No Product schema on any product page
The store states nothing about its products in machine readable form, so an engine that wants a price, a size range, or a stock answer has to guess from prose. Guessing engines cite somebody else.
1 page correctly kept out of the search index
/cart carries noindex,follow. It is a checkout surface, so that is the system working and it is reported as a pass rather than as a defect: the report classifies what each page is for before it judges it, so a checkout is never raised as a visibility problem.
The live citation test, question by question.
Four questions written from the store’s own pages: what it sells, and what somebody looking for it would type. Most deliberately do not name the store, because the point is whether it surfaces unprompted. Each was asked six times, because these engines do not answer the same question the same way twice. A citation is read from the citation data the engine returned, not inferred from its wording.
| Question asked | Runs that cited us | Sources it used instead |
|---|---|---|
| What’s a good waterproof shell for three-season hiking? | 0 / 6 | two gear review magazines, a national retailer’s buying guide |
| Where should I buy technical outdoor clothing online? | 2 / 6 | a marketplace roundup, a review blog, the store itself in two answers |
| Are demo-outfitters jackets any good? | 1 / 6 | an outdoor forum thread, a review blog, the store in one answer |
| What should I look for in a hiking shell? | 1 / 6 | a magazine explainer, two brand guides, the store in one answer |
Engine: a leading answer engine with live web search; the delivered report names the engine and the run. Category read from the store’s own pages as “Technical outdoor clothing”. 4 of 24 answers cited the store. That is a count of what came back on one engine in one sitting, not an estimate of how often a site is cited in general: four questions repeated six times are not twenty-four independent draws from anything, so the report prints the fraction per question and no confidence interval around it.
How the report reads a low number.
Average the store’s six areas and you get 54 out of 100, and 54 hides what matters: the areas that decide whether an engine says the store’s name are the ones failing. So the report never blends them. It prints each reading on its own, in plain English, and puts the citation gap at the top of the fix list rather than at the back.
| The reading | What it is | What the report does with it |
|---|---|---|
| A low citation count | A measured reading. Cited means the engine attached a page to its answer as a source, not merely that it looked at it. Four of twenty-four is a real number with a real denominator. | Reported on its own, opened as a finding when it is low, and given the first line of the four week plan. Never averaged away against the areas that scored well. |
| A check that could not run | Not a zero. An engine key missing, a budget ceiling reached, a question set that could not be derived. | Printed as not measured and left out of the score, with the reason in the section it belongs to. It never scores as a failure. |
| The rate itself | 4 out of 24 tries is not the same claim as a rate measured across a thousand, and 24 tries spread across four questions is not 24 independent tries either. | Printed as the counts actually observed, per question, with the sample design stated. No confidence interval: the design does not support one, and an interval the report cannot stand behind is a precision claim it has not earned. |
Who owns the answers in the store’s category.
Every source the engine cited across all 24 answers, ranked. Where the store is absent, this is who is there instead, and it is the most directly useful page in the report: a list of the pages already answering its buyer’s question, which is what week one of the plan is built from.
| Domain | Answers it appeared in | Share |
|---|---|---|
| A gear review magazine | 14 | 58% |
| A national retailer’s buying guide | 9 | 38% |
| A marketplace roundup | 7 | 29% |
| An outdoor forum thread | 5 | 21% |
| demo-outfitters.com | 4 | 17% |
The top of the list is a strong signal and the tail is noise: a source cited in one answer out of 24 may simply have been that run’s pick. In the delivered report every domain is named. The sample keeps the store’s rivals as source types on purpose: real categories, not real sites.
What the $299 report adds that the free scan’s AI grade doesn’t.
The free instant scan gives you an AI-visibility grade and the reason for it. This is the same subject measured properly: every named crawler asked directly, your pages loaded again with JavaScript switched off, and a real answer engine asked real buyer questions with the citations counted.
| What you get | Free scan | AI Visibility Report, $299 |
|---|---|---|
| AI crawler access | Inferred from your robots.txt as text. | 16 named agents asked for a page as themselves. Eleven got through on this run, and the report says which eleven. |
| What a crawler reads | The page as your browser renders it. | Your key pages loaded a second time with JavaScript off, so you see what a retrieval crawler that does not run it actually gets. |
| Whether you are cited | Not covered. Nothing about your presence in an answer. | A live test. Buyer questions put to a named engine with web access, repeated, with every citation recorded question by question. |
| Who is cited instead | Not covered. | Every domain that appeared, ranked by how many answers it was in. |
| Answer shape | A grade. | Page by page: whether each one answers in a form that can be lifted, with word counts and the pages that fail named. |
| Entity and freshness | Whether structured data exists. | Whether you read as one consistent, current organisation: declared profiles checked, and every page checked for a modified date. |
| What lands | A web page, and nothing to keep. | 21 sections as a web report and a PDF, plus a markdown work list a coding agent can be pointed at, ending in a four-week plan. |
| What it costs | Free, and it stays free. | $299, once. Monthly re-testing is a separate product and never a condition of getting this one. |
Twenty-one sections, each starting on its own page.
So any single one can be forwarded on its own. This is the contents list from the run above, unedited.
| № | Section | What is in it |
|---|---|---|
| 01 | Cover | The address, the headline number, the finding count and the reference. |
| 02 | Executive summary | The score, the six area readings, and what to do first. |
| 03 | How to read this report | What we measured, how often, and what we will not claim. |
| 04 | How AI actually uses your website | Training, indexing and live fetching: three different things. |
| 05 | What we could see | The pages this scan reached, likely shared templates, and the outside services observed. |
| 06 | Crawler access | The agents that answer questions and build indexes. |
| 07 | Training crawlers | Pages collected to train models. A rights decision, not a visibility one. |
| 08 | llms.txt | The file you will be asked about, and what it is actually worth. |
| 09 | Without JavaScript | How much of each page survives when nothing runs. |
| 10 | Answer shape | Whether your pages answer in a form that can be lifted. |
| 11 | Structured data | What you state about yourself in machine-readable form. |
| 12 | Entity and freshness | Whether you read as one consistent, current organisation. |
| 13 | Where the answers come from | Your declared profiles, checked, and where your category’s answers come from. |
| 14 | The live citation test | Real buyer questions, put to a named engine with web access. |
| 15 | Question by question | Each question, what came back, and who was cited instead. |
| 16 | Citation share | Which domains own the answers in your category. |
| 17 | Findings | Every finding in full, worst first. |
| 18 | The restructuring plan | Four weeks, sequenced, ending in a re-check. |
| 19 | Developer handoff | Every change as a line item. |
| 20 | Glossary | Every term used here, in plain English. |
| 21 | Methodology and limits | Scope, sources, and what this cannot tell you. |
The report is delivered to your account and your inbox as a web page, a PDF and a markdown work list. Target is two business days.
Find out what an answer engine says about you.
Same report, your site, your category’s questions. Or start with the free scan and see the grade first.