Before you pay $1,500, see exactly what arrives
Eight sections, real tables, and one worked finding — built from console data that is already published on this site, on our own two properties. Not a mock-up, and not only the flattering half.
Scope and method statement
Every report opens by stating what was measured and what was not, so that no figure later in the document has to be taken on trust. Yours covers 30 prompts across six engines — Google AI Overviews, ChatGPT, Gemini, Claude, Microsoft Copilot, Perplexity — plus up to three named competitors and a content gap analysis. Each recorded answer carries the date it was captured and the country it was run from, because these engines answer differently by market and an undated screenshot proves nothing.
The section also states the limits plainly. Only two of the six engines report anything back to publishers at all: Microsoft Copilot, through Bing Webmaster Tools, and Google, through Search Console — and Google publishes impressions with no citation count. ChatGPT, Gemini, Claude and Perplexity report nothing. Where a number exists it is named and sourced; where it does not, the report says so rather than estimating.
Where you stand today
The baseline. For this sample the subject is HomeCalc.ca, a Canadian real-estate calculator site that was under three months old when the audit began and sitting at near-zero.
Source — counted across two consoles — Bing Webmaster Tools' AI Performance report (Microsoft Copilot) for citations, and Google Search Console's Generative AI features report for impressions. The AI appearances total is 17,800+ citations in Microsoft Copilot (Bing AI Performance) plus 15,000 impressions in Google generative AI features (Search Console, Beta).
Which pages earn citations
A total tells you nothing about what to do. This table is where the report becomes actionable: it names the pages the engines actually pull from, which is almost never the pages a business assumes.
| Page | Type | Citations |
|---|---|---|
| How to Qualify for a Mortgage | Guide | 1.8K |
| Mortgage Affordability Calculator | Tool | 1.5K |
| Closing Cost Calculator | Tool | 1.4K |
| Mortgage Amortization Calculator | Tool | 1.3K |
| Ontario Land Transfer Tax Calculator | Tool | 1.0K |
| Mortgage Qualifier Calculator | Tool | 831 |
Source — measured in Bing Webmaster Tools' AI Performance report (Microsoft Copilot and partners), trailing three-month view.
Query-level citation share
Share matters more than volume. A query where you hold 39% of all citations is a position worth defending; one where you hold 2% of a bigger number is a queue you are standing in.
| Grounding query | Citations | Share |
|---|---|---|
| land transfer tax ontario | 443 | 39% |
| Canada mortgage qualification requirements | 263 | 39% |
| property transfer tax bc | 263 | 29% |
| closing costs | 224 | 18% |
| amortization period | 206 | 21% |
Source — measured in Bing Webmaster Tools' AI Performance report (Microsoft Copilot and partners). Sorted by citations, not by share — a high share on a handful of citations is noise, and the report says so wherever it appears.
Engine-by-engine read
One row per engine, per prompt. In your report each row carries the engine’s answer recorded verbatim, the date, the country it was run from, and which competitor was named if you were not. Below is the shape of it, and what each engine can and cannot tell us.
| Engine | Reports back to publishers? |
|---|---|
| Microsoft Copilot | Yes — citation counts, cited pages and grounding queries, via Bing Webmaster Tools. |
| Google AI Overviews | Partly — impressions only, in Search Console. No clicks, no citation count, no queries. |
| ChatGPT | No — measured by hand, recorded verbatim with date and country. |
| Gemini | No — measured by hand, recorded verbatim with date and country. |
| Claude | No — measured by hand, recorded verbatim with date and country. |
| Perplexity | No — measured by hand, recorded verbatim with date and country. |
This table is the reason the audit is done by a person rather than a dashboard. Four of the six engines publish nothing at all, so any tool claiming a score across all of them is inferring it. The report distinguishes the measured from the observed on every line.
Prioritized findings
Findings are ranked by what they cost you, not by how easy they are to fix. Here is a real one, worked end to end — and it is about this site, not a client’s, because a sample that only shows wins is not a sample.
The domain earns thousands of AI citations. The page it sells earns twelve.
Observed. HamiTahm.com recorded 7,100+ Copilot citations across the window, and the curve was flat across the entire window — no growth, ~91% concentrated on a single page. Meanwhile homecalc.ca— same owner, same window, under three months old — recorded 14,600+ on a curve described as near-zero → steep, sustained growth.
Diagnosis. The volume is real but it is landing in the wrong place. An old essay, “The 10,000-Hour Rule”, absorbs 6,500of them. The commercial page — the one the business actually sells — earns 12.
Commercial impact. Over the same quarter that essay produced 24 Google clicks, and the whole domain produced 44 at an average position of 42.8. Citations are not traffic and traffic is not revenue; the report says which of the three a number belongs to every time it shows one.
What this means for you. A visibility number that goes up is not automatically a win. Before anyone spends a dollar chasing more citations, the report establishes whether the citations already being earned point at anything that makes money — and if they do not, that becomes finding 01 rather than a footnote.
Source — Bing Webmaster Tools → AI Performance (Microsoft Copilot and partners), April 25, 2026 – July 25, 2026, pulled July 27, 2026; Google Search Console for the clicks and position figures. The full dataset is published, licensed CC BY 4.0, at /research/.
Your report contains the full ranked list in this format. Each entry names the pages involved, the evidence, and what changing it is expected to move.
Content gap
The last analytical section, and usually the one that produces the most work for the client’s team. It lists the questions buyers in your category are asking the engines where a competitor is answering and you are not — separated into pages you do not have, pages you have but that are not structured to be extracted from, and pages that exist and are being passed over for a reason the report names.
For the sample subject the pattern is visible in section 03 above: five of the six most-cited pages are tools rather than articles. That is a content-gap finding, not a coincidence, and it is the kind of thing the section exists to surface before you commission another blog post.
What this report does not claim
- It does not promise a citation count. Nobody can, and any report that does is selling you a forecast dressed as a finding.
- It does not include execution. The deliverable is the measurement and the recommendations; your own developer or agency ships the changes.
- It does not report on engines that publish nothing as though they did. Where an answer was observed by hand, the report says so, with the date and the country.
- It does not treat a citation as a customer. Section 06 above exists precisely because those two came apart on our own site.
This, about your site
AI Visibility Audit — $1,500 CAD, CAD, flat fee, one-time. 30 prompts across six engines, up to three competitors, a content gap analysis, a 60-minute walkthrough and 14 days of follow-up. Delivered within 7 business days.
Not ready for the full audit? The audit page covers scope and turnaround in detail.
Sample report — FAQ
Is this a real report or a mock-up?
The data is real. Every figure on this page was read from Bing Webmaster Tools' AI Performance report or Google Search Console, and each one is published elsewhere on this site with its source named. The subject is our own two properties — HomeCalc.ca and hamitahm.com — rather than a client's audit, so nothing here depends on you trusting an anonymisation.
Why use your own sites instead of a client's?
Two reasons. A client's report is theirs, and redacting one still asks you to believe the redaction. And using our own sites let us include the bad result: hamitahm.com is in here as a site that earned thousands of AI citations while the page it actually sells earned twelve. A sample that only shows wins tells you nothing about what happens when the answer is unflattering.
Does the real report show the method?
Your report shows every input and every result: the prompts, the engines, the country each answer was recorded from, and the date. What it does not publish — here or in the report — is how the 30 prompts are selected and how variance between model runs is handled. That is the part you are paying for, and putting it in a public sample would make it worth nothing.
How long is the actual report?
Longer than this page, because yours is specific to you: every prompt, every engine's answer recorded verbatim with its date and country, the competitor comparison, the content gap, and the full prioritized finding list. This page shows the structure and one worked finding, not the volume.
Do I get a call, or just the document?
Both. The $1,500 CAD audit includes the written report, a 60-minute walkthrough call, and 14 days of follow-up access. If you then want the findings turned into a fixed-scope plan your own developer can ship, that is the AI Visibility Action Plan (From $4,500 CAD) and the audit fee is credited toward it.