How to Use Search Console and Bing Webmaster Data With AI Citations to Find Pages That Rank But Never Get Cited (2026)

Key takeaways

• Google Search Console and Bing Webmaster Tools now each have an AI report, but they measure different things. Google counts impressions in AI Overviews and AI Mode. Bing counts citations in Copilot and Bing's AI answers.
• A page that has search impressions and zero Bing citations is your first candidate for a "ranks but never cited" page.
• Neither first-party report covers ChatGPT, Perplexity or Claude. You need manual checks or a prompt-tracking tool for those.
• Compare each uncited page's content type and intent against what actually gets cited. A blog post competing for a commercial query is up against product pages and listicles.
• Treat all of this as diagnosis, not proof. Nobody outside the model can say exactly why one source was picked over another.

Why ranking and being cited are different problems

For twenty years the question was simple: where do I rank? Now there is a second question, and it has a different answer. A page can sit at position four for a keyword, collect steady impressions, and never show up as a source in any AI answer.

Part of the reason is crowding. In Promptwatch's data on unique domains per ChatGPT response (Sep 28 to Oct 4, 2026), 47.6% of responses with citations cited 10 or more domains, and 66.1% cited at least 8. Only 6% cited three or fewer. Ranking well gets you into the pool. It does not guarantee one of the slots.

The other reason is that AI systems don't search the way people do. They generate their own conversational queries, retrieve pages for those, and cite the ones that answer cleanly. Your keyword list and those queries overlap, but they are not the same thing.

So the job is a join between two datasets: what search engines show, and what AI engines cite. This guide walks through doing that with free first-party data, then filling the gaps.

What each data source actually tells you

Before building anything, be clear on what each report measures. Mixing them up is the most common mistake.

Google Search Console: the generative AI performance report

Google announced the Search generative AI performance reports on June 3, 2026, with dedicated views for Search and Discover. Per Google's help page, the report was rolled out to all sites worldwide by August 31, 2026. It shows impressions, pages, countries, devices (Search only) and dates, at hourly, daily, weekly or monthly granularity. It covers AI Overviews and AI Mode and excludes Search Labs experiments.

Two things matter here. First, this data was already inside your overall Performance report. The new view only separates it out. Second, an impression means a URL appeared in a generative AI feature. It is not a count of ChatGPT or Perplexity citations. A third-party read from Agency Dashboard (June 30, 2026) also noted no clicks, CTR or query data in the view, and that AI Overviews, AI Mode and Discover were blended. Check the current interface, since that may have changed.

Bing Webmaster Tools: AI Performance

Microsoft launched AI Performance in public preview in February 2026. It covers Microsoft Copilot, AI-generated summaries in Bing, and select partner integrations. You get Total Citations, Average Cited Pages (unique pages cited per day), grounding queries, page-level citation counts and a visibility trend.

Microsoft's own caveats are worth repeating. Grounding queries are a sample of citation activity, not the full set. Citation counts say nothing about placement, ranking, page importance or authority. It is a citation log, not a traffic report.

On June 16, 2026, Microsoft added four preview capabilities:

• Intents classifies grounding queries into categories like Informational, Commercial, Navigational, Learn and Solve, Research, Creation and Local.
• Topics groups related grounding queries into clusters. "Solar panels" and "residential solar installation" would both map to Solar Energy.
• Citation Share is your citations as a percentage of all citations shown for the same grounding query. It doesn't expose competitor domains.
• Compare overlays a prior period on the current view.

CXL's write-up of the Bing report is a useful reference for what the screens look like and where the report falls short.

A quick side-by-side

No single row answers the question. The workflow below stitches them together.

Step 0: Make sure Bing can see you at all

This is the step people skip. Copilot grounds its answers in Bing's index, so a page that Bingbot can't fetch, or that Bing never indexed, can't be cited, however well it does in Google. Nytro SEO makes this point directly, and adds that sites that have ignored Bing often find a real share of pages missing because of crawl errors or blocked resources that Google tolerated.

1. Sign in to Bing Webmaster Tools. You can import your site from Google Search Console in a couple of clicks.
2. Check index coverage and crawl errors. Fix blocked resources and server errors first.
3. Confirm robots.txt isn't blocking Bingbot.
4. Set up IndexNow so new and updated pages get pushed rather than waiting to be discovered. Microsoft recommends it for freshness.

If you want to check which AI crawlers can reach your site beyond Bing, the free Agentic Readiness Check and the AI crawler URL inspector are quick ways to test it.

Step 1: Build the list of pages that rank

Export page-level data from both search engines.

1. In Search Console, open Performance, set the date range to at least 3 months, and export pages with clicks and impressions.
2. In Bing Webmaster Tools, open Search Performance and export the same page-level view.
3. Merge them in a spreadsheet on URL. Normalise trailing slashes, protocols and parameters first, or the join will silently drop rows.
4. Add a threshold. Something like 100+ impressions over the period keeps you from chasing pages with no real search presence.

You now have a list of pages that search engines actually show. This is the "ranks" half of the equation.

Step 2: Pull the citation list and find the gaps

Now bring in the citation side.

1. In Bing Webmaster Tools, open AI Performance and export or copy the page-level citation counts.
2. Add them as a column next to your merged list.
3. Filter for pages with impressions above your threshold and zero Bing citations.
4. In Search Console's generative AI report, check the same pages. Mark the ones with no AI Overviews or AI Mode impressions either.

The result is three tiers:

Microsoft itself frames the report as a way to spot pages that are "indexed but less frequently cited". That is exactly tier 1 and 2.

One caution. Bing's data is a sample, and the numbers are small on many sites. Zero citations over 30 days might mean zero. Over a quarter, with a decent page, it's a stronger signal. Use the longest window available.

Step 3: Read the grounding queries as a statement of authority

The grounding queries are the most interesting part of the Bing report, and they look nothing like a keyword list. They are the phrasings a model generates, often longer and more conversational. Nytro SEO notes that teams who verify a site for the first time tend to be surprised twice: Copilot is already citing them from a handful of pages, and the grounding queries don't resemble their target keywords.

Do three things with them:

• Group them with Topics. Which topics does Bing associate with your site? That is, roughly, what it thinks you are an authority on. Labels can be broad for niche sites during the preview.
• Check Intents. If your cited pages are mostly Informational and your uncited high-impression pages target Commercial queries, you've found a pattern, not just a list.
• Compare periods with Compare. A topic gaining citations over 30 days tells you where to double down.

The newer Topics Portfolio view in Bing's beta goes further, showing thematic visibility, topic demand and queries where other sites are cited and yours isn't. If you have access, that's where gap-hunting gets quick.

Step 4: Check the uncited page against what does get cited

Now the diagnostic part. For each tier 1 page, take the relevant grounding queries (or, where you don't have any, the page's top search queries rewritten conversationally) and run them in Google, ChatGPT, Copilot and Perplexity. Note which sources appear instead.

CXL is careful about this, and so should you: running the queries manually can reveal content gaps, but it won't prove why an AI system picked one source over another. You are forming hypotheses.

Compare content type and intent

This is where Promptwatch's research is useful. In Promptwatch Data on ChatGPT citations by content type (Sep 28 to Oct 4, 2026), product pages took 22.4% of classified citations, landing pages 16.8%, listicles 13.6%, how-tos 8.7%, general articles 6.1%, comparisons 5.9% and news 5.7%. By prompt type:

So if your uncited page is a 1,500-word blog article ranking for a commercial query, it's competing against formats that take almost 70% of commercial citations. The fix might not be better writing. It might be a different page type, or a comparison table and clear product details inside the article.

Also keep brand-specific prompts in mind. Promptwatch's unique-domains data shows only 21.3% of brand-specific prompts cite 10 or more domains, with 5 the most common count. Fewer slots, so being absent hurts more.

Check freshness

Content type affects how long citations last. According to Promptwatch's content half-life data, news lost half its cited pages within about 36 days, while landing pages and docs stayed cited far longer. If your uncited page is a dated news-style post, being uncited may just be age.

Authority matters in a surprising way too. Promptwatch's domain rank data showed sites with DR 46 to 75 earning almost half of ChatGPT citations in August 2026, while DR 91 to 100 dropped to about 3%. Being a mid-sized site is not the obstacle people assume.

Step 5: Fix the pages, in order of likely payoff

Based on Microsoft's own tips and the diagnostics above, here is what I'd check on each tier 1 page, roughly in this order.

1. Indexing. Is it in Bing's index? Is Bingbot fetching it cleanly? Nothing else matters if not.
2. Direct answers. Does the page answer the question under a clear heading near the top? Microsoft recommends clear headings, tables and FAQ sections.
3. Evidence. Claims need support: named sources, data, authors. Nytro SEO lists cited sources and named authors as authority signals.
4. Format fit. If the query is commercial, does the page behave like a commercial page, with specifics, comparisons and pricing?
5. Freshness. Update stale facts and push the change with IndexNow.
6. Internal links. A couple of internal links from already-cited pages to uncited ones is a cheap test. One of the practitioner write-ups on grounding queries suggests 2 to 3 links is enough.

Nytro SEO's advice on structure is to bring the metadata and structure of uncited pages in line with the cited ones. That is rules-based work you shouldn't do by hand on a large site. Pull the pattern from your top five cited pages (heading structure, summary placement, schema, table use) and apply it systematically.

Change a batch, not everything. Leave a control group of similar uncited pages untouched so you can tell whether the edits did anything.

Step 6: Cover ChatGPT, Perplexity and Claude

Here's the limit of the free approach. Bing covers Copilot and Bing's AI answers. Google covers its own features. Neither tells you whether ChatGPT cites your page, and ChatGPT is where much of the attention is. Bing's index does feed some of these systems, as one Reddit thread notes, but that's indirect.

You have two options. Do it manually: take your tier 1 pages, write 5 to 10 prompts per topic, run them, and log citations in a sheet. This works for 20 pages and falls apart at 200. Or use a tracking tool.

Promptwatch tracks prompts from the real interfaces of ChatGPT, Google AI Overviews, Gemini, Perplexity, Copilot and others, and its Page Tracker lets you watch a specific URL and see which prompts cite it. It also pulls Search Console and Bing data next to AI crawls and citations in Search Insights, which answers the exact question of which pages rank in Google and Bing but never get cited by AI. Its IP-verified crawler logs help separate an access problem (bots can't read the page) from a content problem (they read it and chose someone else).

Other options in the same space include Otterly.AI, which has published its own workflow for mapping Bing grounding queries against tracked prompts, as well as Peec AI and Profound.

For a wider look at what's out there, the GEO software directory at bestgeosoftware.com lists dozens of tools by feature.

A worked example

Say you run a B2B software site. Your merged sheet shows a page, "Best invoicing software for freelancers", with 4,200 impressions in Google and 600 in Bing over three months. Bing's AI Performance shows zero citations for it. Search Console's generative AI view shows a handful of AI Overview impressions.

Bing's Intents report shows your cited pages are almost all Informational, how-to style guides. This page targets a Commercial intent. Running the likely grounding queries in ChatGPT and Copilot shows listicles from review sites and a few vendor product pages taking the citations. Your page is a 900-word opinion piece with no table and no pricing.

Hypothesis: wrong format for the intent. You rebuild it as a structured comparison with a table, pricing, a short FAQ and named author credentials, push it via IndexNow, and monitor it for 6 to 8 weeks. You leave a similar page untouched as a control. That's the loop. It's not glamorous, but it's measurable.

What this method can't tell you

Be honest about the limits, because the data has real ones.

• Bing's grounding queries are a sample, and the citation counts don't show placement or importance.
• Google's report has no citations and, per one third-party read, no clicks or query data in the view. Impressions are not citations.
• Citation Share doesn't expose which competitors are winning.
• Nytro SEO reports there is no API for the Bing report yet. I haven't confirmed that against Microsoft, so check before you plan automation around exports.
• Small samples mislead. CXL found its ten most-cited pages averaged around position five in organic search, but it was clear that a small sample doesn't establish a general ranking relationship.

So treat "ranks but not cited" as a prioritisation filter. It tells you where to look, not what the answer is.

A repeatable monthly routine

1. Export Search Console and Bing page data. Merge and refresh the three-tier sheet.
2. Review Bing's Intents and Topics with Compare set to the previous 30 days.
3. Pick 5 to 10 tier 1 pages. Run grounding-style prompts across ChatGPT, Copilot, Google and Perplexity.
4. Record the content type and intent of whoever gets cited instead.
5. Edit, push through IndexNow, and note the date.
6. Revisit after 6 to 8 weeks and compare against the control pages.

If you'd rather not run this in-house, an SEO and GEO agency can do it for you. 1001 SEO Media runs this kind of search-plus-AI-citation audit for clients as part of its AI search work.

Final word

The useful insight here is simple. Search impressions tell you a page is findable. Citations tell you it's usable as an answer. The pages in the gap between the two are your cheapest wins, because they already have relevance and a crawlable URL. Start with Bing indexing, build the three-tier sheet, and fix tier 1 pages by matching format and intent to what's actually getting cited.