What was measured?
Three B2B software companies, labelled A to C, each with 90 buyer questions that do not name the company. Every question was asked once a day for at least seven days in October 2026 on seven AI surfaces. Days whose run was still filling when the data was pulled were left out, which left 9,327 answers. These are the numbers behind the guide to audits and retainers.
- Appearance: an answer that names the company in its own words, or cites its website.
- Small audit: 15 questions picked at random, five engines with Google counted once, one day. Up to 75 answers.
- Full audit: all 90 questions on the same five engines, one day.
- Full week: the same five-engine measure across every question and every complete day. Both audits are compared with it.
- Draws: each audit was simulated 20,000 times per company with a fixed seed. The leader and source tests used 5,000.
- Spot: one question on one AI surface.
The three companies
| Company | Category | Days tracked | Complete days used | Answers | Appearances | Named in the text | Cited without the name | Share of answers | Lowest and highest day |
|---|---|---|---|---|---|---|---|---|---|
| A | Cybersecurity software | 7 | 5 | 3,066 | 152 | 7 | 145 | 5.0% | 4.3% to 5.6% |
| B | Cybersecurity software | 8 | 6 | 3,756 | 49 | 22 | 27 | 1.3% | 1.0% to 1.4% |
| C | Healthcare practice software | 7 | 4 | 2,505 | 156 | 91 | 65 | 6.2% | 5.4% to 7.0% |
How close did each audit size get?
| Company | Full week | Small audit: middle 90% | Small audit: typical miss | Small audit: within 2 points | Full audit: middle 90% | Full audit: typical miss | Full audit: within 2 points |
|---|---|---|---|---|---|---|---|
| A | 6.7% | 1.5% to 12.3% | 2.6 points | 44.7% | 5.6% to 7.6% | 0.4 points | 100.0% |
| B | 1.6% | 0.0% to 4.1% | 1.1 points | 86.7% | 1.1% to 2.0% | 0.3 points | 100.0% |
| C | 5.7% | 1.3% to 11.0% | 2.3 points | 47.7% | 4.7% to 6.9% | 0.6 points | 99.9% |
Five small audits of company A drawn one after another, with nothing changed between them, read 6.7, 9.3, 6.2, 4.8 and 9.5 percent. The full week read 6.7.
Do two audits of an unchanged brand agree?
| Company | Small audit: gap of 2 points or more | Small audit: 9 in 10 gaps under | Full audit: gap of 2 points or more | Full audit: 9 in 10 gaps under |
|---|---|---|---|---|
| A | 43.6% | 4.1 points | 6.3% | 1.8 points |
| B | 11.9% | 2.7 points | 0.0% | 0.9 points |
| C | 39.9% | 4.0 points | 7.0% | 1.8 points |
What did a small audit get right?
| Company | Named the week's leader | Named the week's top three | Most-cited websites found, of the week's 10 (typical) | Read the brand as absent |
|---|---|---|---|---|
| A | 78.3% | 96.0% | 10 | 1.2% |
| B | 95.7% | 99.2% | 10 | 31.5% |
| C | 69.8% | 57.8% | 9 | 3.4% |
How much moved from day to day?
| Company | Spots with an appearance | Came and went | Every day | One day only | Seen on a single day (median) | Questions with an appearance | On one surface only |
|---|---|---|---|---|---|---|---|
| A | 59 | 50 | 9 | 21 | 49.2% | 32 | 17 |
| B | 19 | 18 | 1 | 7 | 47.4% | 10 | 6 |
| C | 75 | 63 | 12 | 31 | 52.0% | 29 | 11 |
| Company | Websites cited | Cited on one day only | Share cited on one day only | Week's top 20 still in a day's top 20 |
|---|---|---|---|---|
| A | 2,694 | 1,455 | 54.0% | 15 to 18 |
| B | 2,931 | 1,656 | 56.5% | 18 to 20 |
| C | 2,601 | 1,522 | 58.5% | 17 to 19 |
Which AI surfaces brought the companies up?
| Surface | Answers | Appearances | Named in the text | Share of answers | Spots with an appearance | Days present, in those spots |
|---|---|---|---|---|---|---|
| ChatGPT | 1,350 | 153 | 47 | 11.3% | 50 | 63% |
| Microsoft Copilot | 1,350 | 53 | 10 | 3.9% | 18 | 60% |
| Perplexity | 1,260 | 52 | 7 | 4.1% | 30 | 40% |
| Google AI Overviews | 1,318 | 38 | 21 | 2.9% | 21 | 43% |
| Gemini | 1,350 | 36 | 21 | 2.7% | 19 | 45% |
| Google AI Mode | 1,349 | 24 | 13 | 1.8% | 14 | 39% |
| Claude | 1,350 | 1 | 1 | 0.1% | 1 | 25% |
What are the limits?
- Three companies in two categories, tracked for about a week. No company appeared in more than about 7 percent of answers, so the data says nothing about category leaders.
- The answers were collected by a tracking tool, and both audit sizes were simulated by sampling those answers rather than run separately.
- The companies are not named, so the figures cannot be checked against their own websites. Everything else on this page can be checked against the guide.