Home / Sample report
Case 1 · Finished audit · Our own business
What does an AI visibility audit report look like? Here's a real one.
This is the audit we ran on 9 November 2026 on a small Ontario B2B service firm we own. It's our own business, which is why we can publish it. We don't name it here, and we describe the firms the engines picked by type rather than by name. In your report, they'd be named.
We sell this audit to B2B service firms in Ontario for $499. It's the measuring step of answer engine optimization (AEO): your buyers' questions asked in ChatGPT, Microsoft Copilot, Perplexity, Gemini and Claude, with the fixes ranked and walked through on a video call.
The short version
18 of 182
unbranded answers across seven engines named the agency (9.9%).
0 of 56
on questions shaped like the two buyers Copilot had sent us.
2 of 17
buyer searches where it made Bing's first page. Copilot reads Bing's index.
3 of 26
answers from Bing AI Mode, Microsoft's Copilot-powered search answer, named it.
- It wins one kind of question: the kind that describes its exact tools.
Fourteen of its 18 mentions came from three questions: two that described its tool stack (11 of 14 answers, first in 9 of them) and one about connecting an ERP system to AI agents (3 of 7).
- It's missing from the questions real buyers ask.
Questions written in our real leads' own words: 0 of 56 answers. Questions shaped like the two buyers Copilot sent us: 0 of 56.
- Its ERP win came from a page that no longer existed.
One engine cited a service page that had returned 404 since a site rollback. A buyer who clicked would land on an error.
- Bing had it on page 1 for 2 of 17 buyer searches.
On the other 15, page 1 went to other firms, software vendors and Microsoft's own pages. The broad "AI consulting firm" and "implementation partner" searches were led by Clutch and "best of" list pages.
- Bing AI Mode named it 3 times out of 26, first each time.
All three were questions that described its tools. On every question shaped like the Copilot buyers, it recommended someone else.
- Nothing outside its own site vouched for it.
No backlinks, no Google Business Profile, no AI mention records, no Clutch or directory listing. The engines went looking for exactly those things.
- Month over month, flat.
On October's 15 questions and five engines: 5 of 75 now against 8 of 75 in October. That's within run-to-run variation.
How we ran it
The agency gets its leads through its website, and two buyers had told us Copilot pointed them to it. Both came from companies that live in Microsoft 365. Our first audit, on 6 October 2026, found the agency in 8 of 75 answers (10.7%). This second one added Copilot and 12 new questions written for buyers like those two.
| Item | What we did |
|---|---|
| Questions | 27. Fifteen from October's audit (8 written in real leads' own words, 6 generic searches, 1 control) and 12 new ones for buyers like the two Copilot sent: mid-sized firms, Microsoft 365, ERP, a hotel, a first AI project. One question named the agency; it's left out of every count. |
| Engines | Seven. GPT-5.5, GPT-6 Luna, Perplexity sonar-pro, Gemini 3.8 Flash and Claude Sonnet 5 through DataForSEO's AI answer API, web search on. The ChatGPT app, location Canada, through DataForSEO's ChatGPT scraper. And Bing AI Mode for Copilot. |
| Copilot | The answer API has no Copilot option and copilot.microsoft.com needs a sign-in, so we used Microsoft's own surfaces, signed out, in a browser: Bing web results for 19 searches and Bing AI Mode for all 27 questions. |
| Counting | One answer per question per engine. "Named" means the agency's name is in the answer text. Positions read by hand. |
| Raw data | Every answer, its citations and the searches each engine ran on its own, saved to files and a spreadsheet. |
| Tool cost | About US$6 to $7 in API and data fees for the whole run. The work is in writing the right questions and reading the answers. |
Five of the 27 questions, word for word (city removed)
- Which companies in [its city] can build custom AI agents and automations for a mid-sized business?
- Who can build AI agents that work with Microsoft 365, Outlook and Teams for a company in [its city]?
- Best AI consulting firms for small and mid-sized businesses in [its region]
- Recommend an AI implementation partner in [its city] for a company that has never used AI before
- best AI automation agency [its city] for small business
Limits. These results come from engine APIs, the ChatGPT app and signed-out Bing. They're a repeatable proxy, not exactly what a signed-in Copilot user sees (a different model, their Microsoft 365 work data, personalisation), and every engine varies from run to run.
Which AI engines named it?
| Engine | Named | Share of 26 |
|---|---|---|
| GPT-5.5 (API) | 5 of 26 | |
| ChatGPT app (Canada) | 3 of 26 | |
| GPT-6 Luna (API) | 3 of 26 | |
| Bing AI Mode (Copilot) | 3 of 26 | |
| Claude Sonnet 5 | 3 of 26 | |
| Perplexity sonar-pro | 1 of 26 | |
| Gemini 3.8 Flash | 0 of 26 | |
| All seven | 18 of 182 | 9.9% |
Without Bing AI Mode, the six other engines named it in 15 of 156 answers (9.6%).
Did anything change since October?
| Engine | 6 Oct | 9 Nov |
|---|---|---|
| ChatGPT app | 2 | 1 |
| GPT-5.5 | 1 | 2 |
| Perplexity | 2 | 0 |
| Gemini | 0 | 0 |
| Claude | 3 | 2 |
| Total | 8 of 75 | 5 of 75 |
The drop is within run-to-run variation, so we don't read anything into it. That's why a single check isn't a trend, and why the re-check compares the same questions month to month.
Which kinds of questions named it, and which didn't?
| Kind of question | Questions | Answers | Named |
|---|---|---|---|
| Written in our real leads' own words | 8 | 56 | 0 |
| Shaped like the buyers Copilot sent us: a mid-sized firm, Microsoft 365, an implementation partner, a first AI project, a hotel, proposals and quoting, a private document assistant, handing over ownership | 8 | 56 | 0 |
| Short service searches, such as AI automation consultant [its city] | 7 | 49 | 4 |
| An ERP-integration question | 1 | 7 | 3 |
| Two questions describing the agency's exact tool stack | 2 | 14 | 11 |
| All unbranded | 26 | 182 | 18 |
Where it was named, and how high
| Question | Named | Engine and position |
|---|---|---|
| Stack question A (its tools, its city) | 5 of 7 | #1 on ChatGPT app, GPT-5.5, GPT-6 Luna, Bing AI Mode and Claude |
| Stack question B (its tools, its city) | 6 of 7 | #1 on ChatGPT app, GPT-6 Luna, Bing AI Mode and Perplexity; #2 on Claude; #5 on GPT-5.5 |
| An ERP-integration question | 3 of 7 | #1 on GPT-5.5 and GPT-6 Luna; #4 on ChatGPT app |
| A search naming its main tool | 2 of 7 | #1 on Bing AI Mode; #3 on GPT-5.5 |
| best AI automation agency [its city] for small business | 1 of 7 | #4 on Claude |
| AI automation consultant [its city] | 1 of 7 | #3 on GPT-5.5 |
| The other 20 questions | 0 of 140 |
The agency was named first whenever a question described its exact stack. It wasn't named for the problem a buyer has (missed calls, online ordering, payroll, a hotel's back office, proposals), or for the kind of firm a mid-sized company looks for (an implementation partner, an AI consulting firm, someone to build Microsoft 365 agents).
It's worth checking whether your firm has the same split. The questions you'd type about yourself aren't always the ones your buyers type.
Who got named instead?
In a client's report this section names every firm, with the pages the engines cited for each. Here we describe them by type.
| Question | Who got the answer |
|---|---|
| AI agents for Microsoft 365, Outlook and Teams | Microsoft partners |
| Best AI consulting firms for small and mid-sized businesses in [its region] | Five firms. Bing AI Mode gave publishing pricing or timelines as a reason. |
| Short searches on Bing, such as ai consulting firm [its city] and ai implementation partner [its city] | Clutch's listings for the city and "best AI consultants in [its city]" list pages |
| Custom agents for a mid-sized business; build and hand over ownership; a company new to AI (Bing AI Mode) | Other AI agencies and consultancies, 3 to 6 per answer |
| An ERP-integration question (Bing AI Mode) | Five other firms |
One pattern ran through all of them: the engines named firms they could find proof for elsewhere, on directories, lists, partner pages and price pages.
What did the engines say about the agency?
- When it was named, the reason was a paraphrase of its homepage: its city, its tools, who runs it.
- Two engines still described a service from a page that had been removed.
- Asked about the agency by name, all seven engines described it accurately.
One OpenAI model told a buyer it "couldn't confirm detailed client case studies, independent reviews, or public pricing."
The engines only know what your site and the rest of the web tell them. Thin outside proof shows up in the answers.
What did Copilot and Bing show?
- Bing web, signed out: page 1 for 2 of 17 buyer searches, #1 for one and #5 for another, both naming its main tool. Absent from page 1 for ai consulting firm [its city], ai implementation partner [its city], custom ai agents [its city], ai automation consultant [its city] and 11 more.
- Bing AI Mode: 3 of 26, first each time, all three stack questions. On every Copilot-buyer question it recommended someone else: Microsoft partners for Microsoft 365 work, and firms that publish prices and timelines.
- For ai agents for microsoft 365 [its city], Bing's first page was almost entirely Microsoft's own pages.
Read the Copilot findings in full, with the searches, what Copilot recommended question by question, and the Microsoft-side fixes.
Which sources did the engines read?
| Source | Answers | Bar |
|---|---|---|
| Another firm's website (most cited) | 18 | |
| A second firm's website | 17 | |
| A third firm's website | 16 | |
| The agency's own website | 15 | |
| 11 | ||
| Clutch | 5 |
Counts from the written report. The raw summary file differs by one or two answers for some sources, so read these as approximate. Gemini's links run through a Google redirect and aren't counted by domain.
Which of its own pages did the engines use?
Cited or retrieved: its home page 17 times, its About page 9 times, and two removed service pages 4 and 3 times. Both removed pages returned 404.
What did the engines search for on their own?
The OpenAI models ran their own web searches while answering. They included [its city] Microsoft Solutions Partner AI agents Microsoft 365 Teams Outlook consulting and Clutch [its city] AI consulting companies artificial intelligence, plus searches for reviews of the firm and for a personal profile on LinkedIn.
That's the engines telling you what they want: partner status, directory listings, reviews, and a real person behind the firm.
What did the site and listing checks find?
| Check | Result |
|---|---|
| sitemap.xml and llms.txt | Both returned 404, removed in an earlier site rollback |
| robots.txt | No rules and no Sitemap line. Nothing blocked Bing. |
| Removed pages still cited or linked | Three pages, all 404. A directory profile, also 404, still showed on Bing's first page for a search on the firm's name. |
| Backlinks | None found |
| Google Business Profile | None found |
| AI mention records (Google AI, Canada) | 0 |
| Clutch or other directory listing | None |
What should it fix first?
Ranked for Copilot, because two of its recent leads said Copilot sent them.
- Bing Webmaster Tools.
Verify the site, submit a live sitemap and turn on IndexNow. Bing's index is what Copilot reads, and Bing was still holding removed pages.
- Bring back the two removed pages.
They were the only pages that won a mid-sized buyer's question (first on two engines for the ERP question), and engines were sending people to their 404s.
- A Microsoft 365 page, written from real Outlook work.
Both Copilot referrals came from Microsoft 365 companies, and only Microsoft partners were answering those questions.
- Get onto the lists Bing ranks.
Clutch (it needs verified client reviews), the city's "best AI consultants" lists, and other directory profiles.
- Proof the engines can find.
Case studies and reviews from clients who came through AI, with their consent; a personal LinkedIn profile that lists the firm as a current role; a company page; a Google Business Profile and a Bing Places listing as a service-area business.
- Pages for the questions buyers ask.
One page each, only for work it does: hotel operations, proposals and quoting, field-service job costing, "your first AI project", and "you own what we build".
- A starting price.
Bing AI Mode and October's engines picked firms that state prices and timelines. The agency stated a timeline and no price.
- Measure Copilot.
Copilot and ChatGPT app visits usually arrive with no referrer and show as Direct, so add "Microsoft Copilot" and "ChatGPT" to the contact form's "how did you hear" list. Re-run the same 27 questions monthly against this baseline.
What did we do next?
- Done the same day (9 November): the two removed pages came back, and a sitemap and robots.txt went live.
- Waiting: Bing Webmaster Tools verification and the sitemap submission need the owner's Microsoft account.
- Not started: fixes 3 to 8.
We haven't re-run the questions yet, so there's no "after" number to show you. A sample report shows the method, not a result, and we'd rather say so.
What would my firm's report look like?
- Your questions, written with you from your services, your towns and the words your recent leads used.
- Your competitors, named, with the pages the engines cited for each.
- The same engines, Copilot included, and the same site and listing checks.
- Fixes ranked for your firm, walked through with you on a video call.
- Private. Nothing is published.
For a 30-person civil engineering firm in Kitchener, the questions might read: Which engineering firms in Waterloo Region handle stormwater management for site plan approval? or Recommend a structural engineer in Kitchener for a commercial renovation. For an IT provider in London, Ontario: Which managed IT providers in London support Microsoft 365 for a 40-person office? Those are examples of how we'd write them, not results.
$499 for the audit: up to 30 buyer questions, each asked in all seven engines, paid half up front and half when the report is delivered. Optional re-check $99/month. Start with a free 30-minute check: we'll ask a few of your buyers' questions in ChatGPT and Copilot before the call.
Questions about the sample report
Is the business in the sample report real?
Yes. It's a small Ontario B2B service firm we own. That's why we can publish its results, and why we don't name it or the firms the engines recommended instead.
How many questions are in an AI visibility audit?
Up to 30 buyer questions, each asked in all seven engines. The sample used 27: 8 written in real leads' own words, 6 generic searches, 1 control, and 12 new ones for buyers like the two Copilot sent us (one of those named the agency and is left out of the counts). Yours are written with you, from your services, your towns and your recent leads.
What does an AI visibility audit report include?
A summary, the method, results engine by engine and question by question, who got named instead, what the engines said about you, Copilot and Bing results, the sources the engines read, site and listing checks, and the fixes ranked. It's walked through with you on a video call.
Did the fixes work?
We don't know yet. The first fixes went live on 9 November 2026 and we haven't re-run the questions. When we do, the comparison goes against this baseline.
Your turn
Want this report for your firm?
Book a free 30-minute check. We'll ask a few of your buyers' questions in ChatGPT and Copilot before the call and show you what came back. The full audit is $499.