GEO became measurable: Bing's AI citation counter, and what 0 vs 40 taught me
Bing now shows how often AI answers cite your site. On two of my sites: 0 vs 40 citations. That gap explains GEO - here's why, and what to do about it.
GEO - getting your content cited by AI answers - used to be the part of SEO you could only guess at. As of February 2026 you can put a number on it. Bing Webmaster Tools shipped a report called AI Performance, and it shows exactly how often AI answers use your site as a source.
I opened it on two of my projects. One sits at 0 citations. The other has 40 in six weeks. That pair of numbers explains more about GEO than a dozen think-pieces - so here is what the report shows, what the gap means, and what to actually do about it.
What Bing AI Performance actually shows
Launched in public preview on February 10, 2026, the report tells you when your site is pulled into the sources behind Microsoft Copilot answers, Bing’s AI summaries and partner integrations. It also lists grounding queries - the phrases the AI used to retrieve your content.
Why this matters: until now, GEO was measured sideways. You ran manual prompts, paid for third-party trackers, or read your logs and guessed. For the first time a search engine puts the scoreboard on the table - citations, the pages behind them, and the trend.
0 vs 40: two real sites
A fresh personal domain (74 pages, live since March): 0 citations. In Bing it has 3 impressions across six weeks. The engine barely sees it - and Copilot simply mirrors that invisibility.
A mature domain I run marketing for (a domain with history, 245 URLs in the sitemap, systematic SEO): 40 citations in six weeks, a peak of 8 in a single day, and the trend is climbing - the last two weeks delivered 19 citations against 8 in the two before. Copilot reaches for its pages more and more often.
Same operator, same methods, opposite results. The only real variable is how each domain sits in Bing’s index.
Why one site gets cited and the other doesn’t
It isn’t magic, and it isn’t llms.txt. Copilot grounds its answers in what is well-indexed in Bing: index depth, a mass of genuinely useful pages, and a link profile. No base, no citations.
That last factor is blunt: the single recommendation Bing gave my under-cited personal site was “not enough inbound links from quality domains.” The same signal that holds a young site back in classic search holds it back in AI answers - because the AI is drinking from the same well.
What to do about it
Four practical takeaways from the two dashboards:
- Treat the Bing index as Copilot’s knowledge base. Check how many of your pages Bing actually sees before worrying about anything fancier.
- Use IndexNow to speed inclusion. On my sites it pings Bing automatically on every deploy, so new pages don’t wait around to be discovered.
- Links still decide it - here too. Inbound links from quality domains move both classic rankings and AI citations.
- Run AI visibility monthly. You can now show a client a citation count and a trend line instead of promising “we’re working on AI presence.”
GEO stopped being a vibe. For Bing and Copilot, it’s a metric you can open, read, and improve - the same way you’ve always worked the index. If you want the bigger infrastructure picture behind AI - where models run and what that costs - I wrote that up in where to run your AI in Europe.
Open your own report in Bing Webmaster Tools under AI Performance. How many citations do you have?