What is the difference between "consulted sources" and "visited pages"?

Consulted sources are the search results AI read while answering — the domains it saw, with their rank position (its research diet). Visited pages are the pages AI actually opened and browsed, a far stronger signal of trust. Google AI Overviews exposes no browsing data, so it is absent from the Visited tab.

What do the Citations, Prompts and Avg. position columns mean?

Citations is how often a domain appeared across the latest answers; Prompts is how many distinct questions it influenced (breadth); Avg. position is its average search rank when consulted, where lower means AI trusts it more. The best targets combine high prompt breadth with a strong average position.

Why does AI cite forums or reference sites more than my own site?

Generative engines favour sources they read as independent and consensus-driven — community threads, third-party reviews, reference and retailer pages — over first-party marketing pages. Across the projects we track, forums like Reddit are routinely the most-consulted source, which is why presence on the pages AI reads often matters more than perfecting the pages you own.

How do I use this to improve my AI visibility?

Sort by prompt breadth, read the source-type mix, and drill into the top domains to get a per-page-type content brief. Focus on the neutral third-party bucket you do not yet control, earn genuine positive presence there, and track prompts where a pinned competitor is a top source but you are absent. Then confirm the impact with the Mention Ledger.

Does this work across every AI model?

Consulted sources are tracked across every model you have enabled — ChatGPT, Google AI Overviews, Claude, Perplexity, Gemini, Grok, and more. Visited pages are tracked for every model that exposes browsing data; Google AI Overviews is the exception and is excluded from that tab.

Where AI Gets Its Answers: Consulted Sources vs. Visited Pages

1. AI Does Not Answer From Your Website

Classic SEO trained us to think about our own pages: rank the page, win the click. Generative engines broke that model. When ChatGPT, Google AI Overviews, Perplexity, Claude or Gemini answer "what is the best CRM for startups", they run a web search, read a handful of results, and synthesise an answer from whatever sources they trust — which is frequently not the brand's own site.

That means the decisive question in GEO is no longer only "does my page rank?" It is "which pages is AI actually reading, and is my brand present on them?" The AI Sources & Destinations drilldown answers that directly, by recording the web behaviour behind every AI answer to your prompts — on the latest run of each prompt per model in your selected range, and, as everywhere in Citation Analytics, with every number clickable down to the receipt: the actual answer and the actual page.

2. First, Three Words That Are Not Synonyms

GEO tooling throws around "mention", "visibility" and "citation" as if they were the same thing. They are three different events, and this drilldown lives in the last two:

Mention — AI named your brand in the answer text. (That is the Mention Ledger.)
Visibility — your domain appeared in the search results AI consulted while answering. It was in the room.
Citation — AI actually opened and used your URL. It read you.

Sources & Destinations is the map of visibility and citation: which domains were seen, and which were actually opened. Keeping the three apart is the difference between "we ranked" and "we were read."

3. Consulted Sources vs. Visited Pages: Two Tabs, Two Meanings

The panel has a toggle, and the distinction between the two views is the key to reading it correctly.

Tab	What it records	Think of it as
Consulted sources	The search results AI read while answering — the domains it saw, with their rank position.	AI's research diet.
Visited pages	The pages AI actually opened and browsed to compose the answer.	Where AI sends its attention.

Consulted is broader (everything in the running); Visited is the far stronger signal, because opening a page is a vote of trust, not just an impression. A domain that is consulted a lot but rarely visited is a page AI keeps seeing but not trusting enough to open — a different problem from a domain it never sees at all. One caveat to state plainly: Google AI Overviews exposes no browsing data, so it is absent from the Visited tab; Consulted sources cover every model you have enabled.

4. Reading Citations, Prompts and Average Position

Each domain in the table carries three numbers, and together they tell you how structurally important a source is:

Citations — the column's label in the table — counts how often the domain appeared in the results AI consulted: visibility events, in the strict vocabulary of section 2. Raw volume of influence; citation in the narrow sense (AI actually opening the page) lives on the Visited tab.
Prompts — across how many distinct questions it influenced. A domain cited many times across a handful of prompts is concentrated; one cited across many prompts is broad. Breadth usually matters more, because it means the source shapes many buyer conversations, not one.
Avg. position — its average search rank when consulted (lower is better). A low average position means AI treats it as a top, trusted source, not a long-tail one.

Read them as a set. Your ideal target is a domain with high prompt breadth and a strong average position: a source AI trusts, across many of the questions your buyers ask. That is where a single earned mention pays off across your whole funnel.

5. "You", "Pinned", and the Source-Type Mix

Three layers of labelling turn the table from a list into a competitive map.

The "You" badge marks your own domains — your main site and any channels you own (support, docs, community subdomains) — so you can see at a glance how much of AI's source diet is your content. The "Pinned" badge marks a competitor or retailer you have chosen to watch, flagged wherever it appears in your sources so you can track exactly which questions they are winning source-share on.

The source-type mix goes a level up. The top sources are auto-classified into a type — Retail, Editorial, Community, Brand site, Reference — and a mix bar summarises their blend, so you can say things like "36% of the top sources AI reads in this space are retail, 11% community." (Newly seen domains may show without a type until classified.) That single sentence reframes strategy: if a third of AI's top reading in your category is retailer pages, your product listings and marketplace presence are a GEO channel, not just an e-commerce one.

6. The Third-Party Lesson: Sources You Don't Own Often Win

Here is the finding that reframes most GEO strategies: for many brands, the single most-cited source is not their own website. Across the projects we track, community forums like Reddit are routinely the number-one consulted source, and we regularly see a competitor's own site consulted more often than the brand's. Independent sources — forums, video platforms like YouTube, reference sites like Wikipedia, and the review, comparison and retailer pages in a category — frequently out-rank a brand's own domain in what AI reads.

When a community thread outranks your own site as a source, optimising only your website means optimising the source AI trusts least.

This is simply how generative engines behave. They favour sources they read as independent and consensus-driven over first-party marketing pages. The practical consequence is blunt: presence on the pages AI reads matters more than perfection on the pages you own. Your own site is table stakes; being well-represented and correctly described on the communities, videos, reference and retailer pages in your category is the actual battle. Run the drilldown on your own prompts and you will see your specific version of this pattern — the exact third-party domains that shape the answers about your market.

7. A Source-First GEO Playbook

The drilldown turns "become more visible in AI" from a slogan into a target list:

Rank your sources by prompt breadth. Sort by the Prompts column, not just Citations. The domains influencing the most distinct questions are your priority targets.
Read the mix bar, then the three buckets. Owned ("You"), competitor/retailer ("Pinned"), and neutral third parties. If retail or community dominate the mix, that is where the work is — and it is rarely your own homepage.
Drill into a domain to get the content brief. Click any source and it breaks down into the exact pages AI used, bucketed by type (support, product, guide, editorial). That tells you precisely which kind of page wins on that domain — a ready-made brief for the content or listing you need there.
Earn genuine presence on the top neutral sources. Helpful community answers, accurate review-site and retailer listings, up-to-date reference pages. Not spam — AI discounts obvious manipulation.
Watch consulted-but-not-visited domains, and your own "You" share. A low "You" share of visited pages means your site is not in AI's reading list for your own category. The fix is two-sided: improve the pages AI already reads, and build the citable equivalents on your own domain so it earns a place in the list.
Track where competitors win source-share. Pin your key rivals and retailers, then look for prompts where a pinned domain is a top source and you are absent. That is a precise, defensible target — same prompt, same source, your brand added.
Close the loop. Pair this with the Mention Ledger to see whether new source presence lifts your answer coverage, and the Sub-Brand Portfolio to make sure the mentions you earn name your products, not just your company.

8. Frequently Asked Questions

Q: What is the difference between "consulted sources" and "visited pages"?: A: Consulted sources are the search results AI read while answering — the domains it saw, with their rank position (its research diet). Visited pages are the pages AI actually opened and browsed, a far stronger signal of trust. Google AI Overviews exposes no browsing data, so it is absent from the Visited tab.
Q: What do the Citations, Prompts and Avg. position columns mean?: A: Citations counts how often a domain appeared in the results AI consulted (visibility events, in the strict sense — citation in the narrow sense lives on the Visited tab); Prompts is how many distinct questions it influenced (breadth); Avg. position is its average search rank when consulted, where lower means AI trusts it more. The best targets combine high prompt breadth with a strong average position.
Q: What do the "You" and "Pinned" badges and the source-type mix mean?: A: "You" marks your own domains and channels; "Pinned" marks a competitor or retailer you chose to watch. Beyond that, the top sources are auto-classified into a type — Retail, Editorial, Community, Brand site, Reference — and a mix bar summarises their blend, e.g. "36% of the top sources are retail, 11% community", so you can see what kind of sources shape your category. Newly seen domains may show without a type until classified.
Q: Why does AI cite forums or reference sites more than my own site?: A: Generative engines favour sources they read as independent and consensus-driven — community threads, third-party reviews, reference and retailer pages — over first-party marketing pages. Across the projects we track, forums like Reddit are routinely the most-consulted source, which is why presence on the pages AI reads often matters more than perfecting the pages you own.
Q: How do I use this to improve my AI visibility?: A: Sort by prompt breadth, read the source-type mix, and drill into the top domains to get a per-page-type content brief. Focus on the neutral third-party bucket you do not yet control, earn genuine positive presence there, and track prompts where a pinned competitor is a top source but you are absent. Then confirm the impact with the Mention Ledger.
Q: Does this work across every AI model?: A: Consulted sources are tracked across every model you have enabled — ChatGPT, Google AI Overviews, Claude, Perplexity, Gemini, Grok, and more. Visited pages are tracked for every model that exposes browsing data; Google AI Overviews is the exception and is excluded from that tab.

Key Takeaways

AI answers from the web, not your homepage: the decisive question is which pages AI reads, and whether you are on them.
Mention, visibility, citation are three events: named in the text, seen in the results, actually opened. This drilldown maps the last two.
Consulted vs. visited: consulted is AI's research diet; visited is where it sends its attention — the stronger signal.
"You", "Pinned" and the type mix turn the table into a map: own content vs. watched competitors vs. a Retail/Editorial/Community/Reference blend.
Third parties often win: forums, reviews and retailers routinely out-cite a brand's own site — drill a domain for the per-page-type content brief.

Sources

Ayzeo platform data (July 2026) — aggregated consulted-source distributions across tracked projects; community platforms rank among the most-consulted domains, and competitor-owned sites are regularly consulted more often than the tracked brand's own site.
Aggarwal, P., Murahari, V., Rajpurohit, T., Kalyan, A., Narasimhan, K., & Deshpande, A. (2024). GEO: Generative Engine Optimization. KDD '24 (arXiv:2311.09735) — on how generative engines select and weight sources when composing answers.
Reuters (2024). Reddit in AI content licensing deal with Google — on the structural role of community content in AI answers.
BrightEdge Research (2025). ChatGPT Brand Mentions vs. Citations: What Triggers Visibility.