For most of 2026, "AI visibility" work for NYC small businesses has meant a lot of guessing: tweak a page, ask ChatGPT a few questions, and hope your brand shows up. On June 16, 2026, Bing changed that by adding Citation Share to Bing Webmaster Tools' AI Performance report — the first time a major platform has shown publishers the actual percentage of citations they own for a given query, not just a raw count.
The launch landed in the same week as a less flattering data point: an Ahrefs analysis of 137,000 sites, published June 15, 2026, found that 97% of llms.txt files received zero requests in May 2026. If your GEO checklist this quarter was "write an llms.txt and wait," it's worth knowing what the data actually says before you spend more time on it.
What Citation Share Actually Measures
Citation Share is the headline of a four-part rollout — Intents, Topics, Citation Share, and Compare — now live in preview worldwide inside Bing Webmaster Tools, free for any verified site.
- Citation Share — your percentage of all citations for a specific query. If an AI answer cites 10 sources and 3 are yours, that's a 30% Citation Share for that query.
- Intents — every grounding query gets classified (Informational, Commercial, Navigational, Local, Research, and more), so you can see which intent categories you actually win.
- Topics — groups queries thematically, useful for spotting which service lines are getting cited and which are invisible.
- Compare — overlays a prior period on the current one, so you can isolate whether a specific content change actually moved your citation numbers.
The Inconvenient Data That Landed the Same Week
Ahrefs' study is the largest llms.txt dataset published to date. Among the findings: only 28% of the 137,000 sites analyzed had an llms.txt file at all, and of the small share that did get requests, 96% of that traffic was bots — only about 4% was human, usually an SEO checking their own link. Named AI tools accounted for just 19.5% of llms.txt fetches, led by GPTBot and Claude-Code.
That doesn't mean llms.txt is harmful to keep. It means it is not the lever that moves AI citation visibility, and adoption so far looks driven by speculation rather than confirmed demand from AI platforms. If you've been polishing an elaborate llms.txt file instead of fixing crawler access or page structure, the data says you've been optimizing the wrong thing.
What to Actually Do This Week
- Verify your site in Bing Webmaster Tools (free, about 10 minutes if you haven't already) and check Citation Share for your 10-20 highest-value service and brand queries. This is your baseline.
- Make one content change, then use Compare. Add FAQ schema or a named-author byline to one service page, wait two to three weeks, and overlay the periods to see whether your Citation Share actually moved — real before-and-after evidence instead of a guess.
- Stop over-engineering llms.txt. Keep a short, accurate one if you have it, but redirect the time you'd spend perfecting it toward confirming AI crawlers (GPTBot, PerplexityBot, Google-Extended) can actually reach your pages, and toward clear on-page structure — headings, FAQs, and schema markup.
- Track Citation Share alongside Google's AI performance data in Search Console. Comparing the same month across Bing/Copilot and Google AI Overviews gives you a fuller picture than either number alone, especially if you're already split between ChatGPT and Perplexity citation strategies.
Why This Matters Now
Most NYC business owners we talk to already believe they should be doing GEO work — they just haven't had a number to act on. Citation Share is the first concrete, free one a major platform has handed publishers. Pair it with an honest look at where llms.txt sits on your priority list, and this week's update becomes a real measurement system instead of another item on a checklist. Apex Zone's GEO team can verify your Bing Webmaster Tools setup, baseline your Citation Share, and build the content plan to move it — starting this month.