A real AI visibility scorecard, explained
This is the free one-page scorecard we produced for an Edinburgh estate agency from our 24 October 2026 check. We've replaced the agency, its competitors and anything else that could identify them, and updated the price box. Otherwise the findings are unchanged.
How to read it
- Headline score: 45/100, likely range 35–59. This combines how often the agency was named, how high up and whether its own website was cited. The range matters more than the number. Moves of under about 10 points are noise.
- Share of voice: 28%. Of all the times these four agencies were named, Agent B accounted for 28%. A city-wide brand took 47%.
- Which assistants name you. ChatGPT named Agent B in 4 of 16 answers and Google AI Mode in 5 of 16. On its own area's questions it was named every time: 2nd on ChatGPT and 1st on Google. Assistants we couldn't measure are marked as such, not guessed.
- The basics. The website was blocking some AI crawlers, it had no estate-agent markup, one portal showed the wrong street number, and reviews were thin. Each item is rated red, amber or green, and anything unverified is marked grey.
- Three fixes. These are plain-English actions, such as adjusting a firewall setting that takes about 15 minutes.
Free scorecard vs the full report
| Free scorecard | AI Visibility Report (£195) | Monthly tracking (from £59) | |
|---|---|---|---|
| Questions | 8 | 25+, tailored to you | Your tracking set, monthly |
| Runs per question | 2 per assistant | Several per assistant | Several per assistant |
| Assistants | ChatGPT and Google AI Mode, with others added as they become measurable | ||
| Competitors | 3 local | Your local set | Your local set, with trend |
| Sources the AI relied on | Summary | Full breakdown | Changes month to month |
| Basics check | Key items | Full check | Re-checked |
| Fixes | Top 3 | Prioritised plan | Done for you on Grow |
Every tier shows scores with ranges and none guarantees rankings or mentions.
Method and limits
We put everyday questions to the logged-out web versions of ChatGPT and Google AI Mode from one location in Edinburgh, and recorded which businesses were named, in what order and which sources were cited. The score weights mention rate most heavily, then position, then own-site citations. Ranges are 95% confidence intervals on the mention rate.
The limits are real. AI answers vary between users, locations and days, and the models change. The free scorecard is a small sample. We check public pages only, and anything we can't confirm is marked "unverified". When we checked, Gemini, Perplexity, Claude and Copilot couldn't be measured this way, so the scorecard says so.