“Are We Showing Up More In The AI Answers?”
The question came up in a monthly review, and what I had ready was a count: how many times our pages had been cited in AI answers. It was higher than last month.
Whether higher was good, nobody in the room could say.
That is what a number with no denominator does to you. It moves, and you cannot say what the movement means — up could be your work, or up could be the weather.
Have a look at your own deck. If the AI visibility slide is a count of citations on a line chart, it has the same hole in it.
Microsoft put the missing half on the table on 16 June 2026, free, inside Bing Webmaster Tools.
How Do You Measure How Much Of The AI Answer Is Yours?
You divide your citations by every citation that appeared for that query.
That is what Citation Share in Bing Webmaster Tools now returns — free, in preview (Bing Blogs, 16 June 2026).
The arithmetic is not the interesting part. Your citations, over the total citations from all sites on that query. Anyone could have written that formula on a whiteboard a year ago.
What changed is that the platform started publishing the denominator. You can read your own logs until the end of time and you will never learn how many times somebody else got cited on the same question.
This is the first major platform to disclose citation share as an official figure. So what does swapping counts for shares actually do to the report?
What A Share Tells You That A Count Cannot
A count moves whether or not you do anything.
If the assistant decides to attach eight sources to an answer instead of four, every site in that answer sees its count go up together. Nobody earned that. If it tightens up again next quarter, everybody’s count falls together, and again nobody did anything.
It is the tide lifting every boat in the harbour, being reported as improved seamanship.
A share survives that, because the size of the pool is in the denominator.
Which means the interesting case is now legible: if the citation pool doubled and your share halved, the extra citations in your count were not yours in any meaningful sense. A line chart of counts cannot tell you that, because a pool that doubled and a pool that held still draw the same line.
Move the internal baseline from count to share and “we went up” starts arriving with something behind it.
The Other Three Views Decide Which Questions You Measure
Citation Share shipped alongside three others: Intents, Topics and Compare.
Intents shows the search queries the AI issues internally while assembling an answer, sorted by what the person was after — informational, commercial, navigational.
Topics groups those queries into themes, so you read subject areas rather than individual keywords.
Compare lays a previous period over the current one and gives you the difference.
None of the three reports a share. They tell you which slice of demand to point Citation Share at, which matters because your share on commercial-intent queries and your share on informational ones call for completely different work.
One caveat Microsoft raises itself: the Topics labels get coarse in specialist fields. This is a preview, and that part should be read with a discount.
Don’t Read The 1 June Jump As Your Campaign Working
Impressions in the AI Performance report — the screen that tallies how often you appear and get cited in AI answers — jumped sharply from 1 June 2026.
Actual citations did not increase. The figures were backfilled: historical data written into the record retrospectively, as Microsoft’s Krishna Madhavan explained on X in early July 2026 (PinMeTo).
Backfill, if the word is new, just means the past got topped up after the fact.
Anyone who started a campaign in June is going to misread this. That step in your chart — are you sure it was you?
I would have reported it as last month’s work paying off, without hesitating. The dashboard puts no note on the screen telling you otherwise.
Two things fix it. Compare like-for-like periods, both of them after the backfill. And mark 1 June on your internal charts as a measurement event, the way you would annotate a tracking change.
Separating what your work did from what your instrumentation did is what decides whether anyone in the company trusts the AI visibility numbers next quarter. A chart with an unexplained step in it stops getting opened.
How Far This Number Reaches
Citation Share covers Copilot and Bing’s AI answers. What ChatGPT, Gemini or Claude are citing is not in here at all.
Microsoft is explicit that this is an observational metric — not a ranking, not a competitive scoreboard. You cannot see competitor domains, and it is not a share of traffic.
Behaviour on non-English queries is not documented either.
So what Bing Webmaster Tools gives you is your own share, tracked consistently, on one platform. That is the whole of it.
It is still a step up from a count with nothing underneath it. For free, today, that is a perfectly useful thing to have.
Move The Baseline From A Count To A Share
One line on one slide changes.
If the number your company treats as its AI visibility baseline is a count of citations, make it a share. Bing Webmaster Tools costs nothing, so the work here is thirty minutes of reformatting a report.
While you are in there, annotate the 1 June backfill. Line up your comparison windows so both sit after it, and the jump stops getting mistaken for a result.
Then count the rest of the world separately.
The share you can see covers Copilot and Bing answers only, and how ChatGPT or Gemini describe your company will never show up on that screen.
Which means measuring across several models stays your own job, the same as it was before.
Sources
- Microsoft, “New AI Visibility Insights in Bing Webmaster Tools: Intents, Topics, Citation Share, Compare”, Bing Blogs, 16 June 2026 (announcement of four new views in Bing Webmaster Tools, free and in preview. Citation Share reports a site’s citations as a proportion of all citations appearing for a query. Intents classifies the internal search queries the AI issues by informational, commercial and navigational intent; Topics groups those queries into subject areas; Compare overlays a previous period. Microsoft positions the figures as observational metrics rather than a ranking or competitive scoreboard: competitor domains are not visible and the share is of citations, not traffic. Coverage is Copilot and Bing AI answers. Microsoft notes that Topics labels can be coarse in specialist domains.)
- PinMeTo, “Bing’s AI citation surge was backfill, not new visibility”, 2026 (reports that impressions in the Bing Webmaster Tools AI Performance report rose sharply from 1 June 2026 as a result of historical data being backfilled rather than an increase in actual citations, citing Microsoft’s Krishna Madhavan on X in early July 2026.)
