Fusion Insights

Overview

Most AEO tools can tell you which domains an AI engine cites. They stop there — they can't connect a citation to what it actually did for your business. Fusion Insights closes that loop. It joins three things AEO Optima already tracks, using your own page URL as the common key:

1. AI cites page X — from your LLM snapshot citations; from Google Search Console's first-party AI-appearance data (how often your own pages show up in Google's AI Overviews / AI Mode — free, no third-party tool); and, when a DataForSEO connector is configured, competitor AI Overview / AI Mode citation share.
2. Page X gets traffic — from Google Analytics 4, filtered to AI-referral traffic, broken down by landing page.
3. Page X ranks — from Google Search Console, by page.

The result is a per-page table and a set of plain-language insights that answer questions like "which of our pages does AI cite, and what traffic did that drive?" and "where do AI citations and Google rankings disagree?"

The join key: your own page URL

A citation to https://yourbrand.com/blog/guide, a GA4 landing page /blog/guide, and a Search Console page https://yourbrand.com/blog/guide are all the same page. Fusion normalizes every URL to a canonical path — dropping the scheme, host, query string, and fragment, lowercasing, and removing a trailing slash — so the three data sources line up. Only your own pages (those whose domain matches your project's website) enter the fused table; third-party citations are counted for share metrics but never attributed traffic.

Plug-and-play: it only shows what you've connected

Fusion Insights degrades gracefully. You never see a broken or empty panel because a source isn't connected.

If a source isn't connected, its insights simply don't appear and the response notes why (so the interface can offer a "connect" prompt instead of a blank space). Connect it later and those insights light up automatically — no configuration, no migration.

The insights

Cited page → traffic (the flagship loop)

For each of your top pages that AI cites and that has measurable AI-referral traffic, Fusion reports the citation count alongside the sessions, conversions, and revenue that page earned. This is the working loop: AI cited you → people arrived → some converted.

AI-vs-SEO divergence

AI citation and Google rank overlap surprisingly little. Fusion flags both failure modes: pages AI cites that rank poorly on Google (AI sees authority the ranking doesn't), and pages that rank in Google's top 10 but AI never cites (SEO winners invisible to AI). These are two distinct channels — winning one does not win the other.

Visibility leads traffic

Using a lagged correlation across the daily series, Fusion estimates whether your AI visibility predicts your AI-referral traffic, and by how many days. When the signal is statistically significant, you get an early-warning indicator: a visibility dip today forecasts a traffic dip in ~N days.

AI visibility from Search Console (requires Google Search Console)

Google's own first-party measure of how often your pages appeared in its AI Overviews and AI Mode — with impressions, clicks, and CTR per page. Because it comes straight from Search Console (via the searchAppearance data Google launched in June 2026), it's free and needs no third-party connector: connect GSC and your owned-page AI visibility lights up. Note the distinction from the DataForSEO insight below — Search Console reports only your own property's appearance, so it answers "is Google's AI surfacing my pages?", while DataForSEO answers "who wins the AI citation share, including competitors?".

AI Overview citation share & the AIO-vs-organic gap (requires DataForSEO)

The share of AI Overview citations on your tracked queries that point to your own pages versus competitors', and how many of your AI-Overview-cited pages are not in Google's organic top 10 — the gap between winning the AI answer and ranking in the blue links. This is the competitive view Search Console can't provide (it sees only your property), so DataForSEO remains the source for mention share in AI answers.

Citation gap with demand (requires Search Console)

Pages that have real Google search demand (impressions) and rank, but that AI never cites. The audience is searching and the page ranks — making it more citable is the highest-value AEO opportunity.

AI conversion value (requires GA4)

Whether your AI-cited pages convert above your AI-traffic baseline. AI traffic tends to be higher-intent (industry data suggests 4–5× organic), and this quantifies the premium for your own pages with the conversion lift and attributed revenue.

Branded search lift (requires Search Console)

Whether AI visibility tracks branded Google search demand. When your brand appears more in AI answers, branded searches often follow — a second payoff from AI visibility beyond direct citations. Reported only when the correlation is statistically significant.

Cited page at risk (always available)

A page AI currently cites whose citation volume has dropped sharply versus the prior equal-length window — an early warning before the page falls out of AI answers entirely, so you can act on freshness or competitive displacement before the citations disappear.

Social mention volume (requires a social connector)

How often, and where, your brand is mentioned across community sources (GitHub, Hacker News, Reddit, Stack Exchange, and similar). Because large language models train on and cite community discussion, this is a leading indicator of AI visibility — distinct from being cited inside an AI answer. It's a brand-level signal (external posts about you), and when one of those posts links directly to one of your own pages, that's called out too.

The math, briefly

• Lead/lag uses a cross-correlation of daily AI visibility (your brand-mention rate) against daily AI-referral sessions, scanning lags of 1–14 days. Lags are counted in calendar days, and each day is paired only when both series actually captured that day — missing days are left out, never filled with a fake 0%. Because daily series are strongly autocorrelated (today looks like yesterday), the significance test uses the autocorrelation-corrected effective sample size, not the raw day count, so a smooth trend can't masquerade as significance. And because we scan 14 lags, we control the whole scan with a Benjamini-Hochberg false-discovery-rate correction rather than accepting the single best lag at p < 0.05 — a lead is reported only if it survives that correction. It needs at least 14 overlapping daily data points. Branded-search lift uses the same autocorrelation-corrected test. One honest limitation on the traffic side: GA4 only records days that had AI-referral sessions, so a day with genuinely zero AI traffic looks the same as a day GA4 simply didn't report — the correlation is therefore computed over days that do have observed traffic, not an imputed zero on every silent day.
• Citation counts are windowed. Every "in this period" citation number counts the actual citations that occurred inside the selected window (recounted from the raw snapshots), not a page's running lifetime total. This is what makes the cited-page-at-risk drop comparison honest: the current window and the prior window are counted the same way, so a real decline shows up and a page that's simply old doesn't look like it's collapsing.
• Rank uses the same impressions-weighted average position as the rest of the platform (matching Google's own "average position"), so a high-traffic query dominates the average more than a no-traffic one.
• Owned-page detection compares the cited domain's registrable core against your project's website — so blog.yourbrand.com counts as yours, but it errs toward not flagging a page as owned rather than wrongly claiming a third-party page.

Numbers are read from already-captured data — Fusion never triggers a new paid capture on its own. (Capturing fresh AI Overview data via DataForSEO is a separate, metered action.)

Where it appears

Fusion extends surfaces you already use — there is no separate page to learn:

• Analytics — the full section: the dual-axis chart (AI visibility vs real traffic), the fused insight cards, and the per-page fusion table. On a day with no snapshot capture, AI visibility is shown as a gap (a bridged line), not as 0 — the line only drops when your brand was actually less visible, never because monitoring didn't run that day.
• Citations — a focused "Cited Pages → Real Traffic" view: which of your cited pages earn traffic and how they rank.
• Revenue — a focused "AI Citations → Conversions" view: which cited pages drove the converting AI traffic.
• Insights → Discover — fused insights flow into the standard Insights lifecycle, so you can convert any of them into a tracked action.

(Fusion only appears on pages where the page-URL join genuinely adds something — it isn't bolted onto surfaces where it would be noise.)

Measuring impact (the efficacy loop)

When you convert a fusion insight into an action and complete it, the platform later measures whether it actually worked — at the page level. For a page-specific fusion action (e.g. "strengthen /blog/guide"), the action's verification compares that page's AI-referral sessions, Google average position, and AI citations in an equal window after completion versus before. The action detail then shows, for example, "/blog/guide: position 14 → 6, +40 AI-referral sessions" — closing the loop from insight to action to measured outcome. Metrics whose source isn't connected are simply omitted.