What tools do I need to measure GEO vs SEO overlap?

You need a SERP API to capture Google top 10 organic results at scale, such as Apify, and access to the AI engines your buyers use, typically ChatGPT and Perplexity, to collect cited domains per query. A spreadsheet is enough to store per-query domain lists and compute overlap. None of this requires a heavy platform to start, though tools help once the prompt set grows.

How many queries do I need for a reliable GEO vs SEO measurement?

There is no fixed minimum, but a few dozen real buyer queries across your core categories is a reasonable start, and a couple of hundred gives steadier per-query averages. Treat any single run as a directional snapshot, not a verdict, because AI answers shift by run, account, region and phrasing. Consistency of method across runs matters more than raw sample size.

How to Measure Your GEO vs SEO Overlap (the Honest Way)

Q: How do you measure GEO vs SEO overlap?

Build a prompt set of real buyer queries, then for each query collect Google's top 10 organic domains (via a SERP API like Apify) and the domains AI cites in ChatGPT and Perplexity. Match the two lists per query, not in aggregate, and compute the share of AI-cited domains that also rank in Google for that same query. Average those per-query shares. In my benchmark that average was about 22%, even though 165 of roughly 500 domains overlapped at the aggregate level.

Q: Why are domain-level overlap counts misleading?

Aggregate counts pool every domain across every query, so a domain that ranks for query A and is AI-cited for an unrelated query B still counts as overlap, even though those are different searches. That inflates the apparent alignment. In my data the aggregate suggested 165 shared domains, but per-query matching showed only about 22% of AI-cited domains actually rank in Google for the same query. Per-query is the honest metric.

To measure your GEO vs SEO overlap honestly, match sources query by query rather than pooling them. For each real buyer query, collect Google’s top 10 organic domains and the domains AI cites, then compute what share of AI-cited domains also rank in Google for that same query, and average those shares. I am Andrii Byzov, an AI-native fractional CMO and generative engine optimization (GEO) consultant for B2B SaaS. In my AI Search Visibility Benchmark the aggregate looked like 165 of roughly 500 domains overlapping, but the per-query average was only about 22%. The aggregate count overstated alignment. Per-query matching is the method that tells the truth.

Key takeaways

Match per query, not in aggregate. Pooling domains across queries inflates apparent overlap.
In my data, aggregate overlap was 165 of ~500 domains, but per-query overlap averaged only about 22%.
Build the prompt set from real buyer queries, not keyword-tool head terms.
Collect both sides per query: Google top 10 via a SERP API and AI citations from ChatGPT and Perplexity.
Report the AI-only set, the Google-only set, and AI Overview presence alongside the overlap number.
Treat any run as a directional snapshot, given the geo, sample size and engine limitations below.

Why domain-level counts mislead

This is the trap, so it goes first. The tempting metric is a single big number: how many unique domains show up in both your Google rankings and your AI citations across the whole study. In my benchmark that count was 165 shared domains out of roughly 500. It sounds like meaningful alignment.

It is not, really. An aggregate count pools every domain across every query. So a domain that ranks in Google for query A and gets AI-cited for an unrelated query B still lands in the overlap bucket, even though no single buyer search ever saw both. The number answers “do these domains ever appear on both surfaces anywhere” when the question you actually care about is “for this query, does what Google ranks match what AI cites.” Those are different questions, and only the second one maps to a buyer’s experience.

When I switched to per-query matching, the honest figure dropped to about 22%. That gap, 165 domains of apparent overlap versus 22% real per-query overlap, is the whole reason to measure carefully. If you report the aggregate, you will likely overstate how much your SEO work carries into AI answers.

Step by step: how to measure it

1. Build a prompt set of real buyer queries

Start with the questions your buyers actually type, not head keywords from a volume tool. Pull them from sales calls, support tickets, your search console, and the “people also ask” patterns in your category. Aim for a few dozen to start, spread across the categories you want to win. Phrase them the way a buyer would ask an assistant, in natural language, because that is what AI engines respond to.

2. Collect Google’s top 10 per query

For each query, capture the top 10 organic results and record the domains. At any scale beyond a handful of queries, automate this with a SERP API such as Apify, which returns structured organic results you can store cleanly. Decide up front what you exclude, typically ads, maps packs and other non-organic blocks, and keep that rule consistent. Note separately whether an AI Overview appeared, since that is its own signal worth tracking.

3. Collect AI citations per query

Run the same queries through the AI engines your buyers use, in my case ChatGPT and Perplexity, and record the domains each one cites. Capture the cited source list, not just whether you were mentioned. Keep the engine list fixed across runs so results stay comparable. My companion post, how to track AI visibility, covers the cadence and logging discipline for this side in more depth.

4. Match per query, then compute overlap

This is the step that separates an honest measurement from a vanity one. For each query, take the set of AI-cited domains and the set of Google top-10 domains, and find the intersection for that query only. Then compute the per-query overlap as the share of that query’s AI-cited domains that also rank in Google for the same query. Average those per-query shares across the whole prompt set. That average, about 22% in my benchmark, is your headline metric.

5. Pull the AI-only and Google-only sets

The non-overlap is where the insight lives. For each query, list the domains AI cites that Google does not rank (your GEO opportunity surface) and the domains Google ranks that AI never cites (organic strength not translating to AI). Across my sample, established category brands turned up in the AI-only set despite not ranking, which directionally suggests entity authority drives citations differently than on-page SEO drives rankings. For the fuller read on what that AI-only set looked like, see GEO is not SEO.

6. Report AI Overview presence

Track, per query, whether Google showed an AI Overview at all. It is a useful third column because it hints at where Google itself is shifting toward generated answers, and it can change over time independently of both your rankings and your AI citations.

Honest limitations

Any version of this method is a directional snapshot, not a law, so state the caveats plainly:

Snapshot in time. SERPs and AI answers both shift between runs. One capture is a sample, not a verdict.
Geo and personalization. Results vary by region, account and history. Mine reflect a US context; yours may differ.
Sample size. A few dozen queries shows direction; it does not give tight precision. More queries steady the average.
Engine choice. Covering ChatGPT and Perplexity is not every assistant. Add or swap engines to match your buyers, but keep the set fixed across runs.
Matching rules. Your exclusions (ads, maps, subdomains, redirects) affect the number. Document them and apply them the same way every time.

The point of naming these is not to undercut the method. It is that a measurement you can defend, with its limits on the label, beats a tidy vanity figure you cannot.

What to do with the result

A low per-query overlap, like my 22%, is the practical case for treating GEO as its own discipline rather than a byproduct of SEO. If most of what AI cites for a query is not what Google ranks for that query, then optimizing only for Google leaves most of the AI answer untouched. Use the AI-only set as your GEO target list and the Google-only set as a reminder of organic strength that is not yet translating.

This is the work I do as a consultant. If you want help running the measurement and acting on it, AI search optimization and the AI visibility score are where I turn findings like these into a plan. I also write more about method on LinkedIn.

FAQ

How do you measure GEO vs SEO overlap? Build a prompt set of real buyer queries, collect Google’s top 10 (via a SERP API like Apify) and AI citations (ChatGPT and Perplexity) per query, match per query, and average the per-query overlap. In my benchmark that averaged about 22%, despite 165 of roughly 500 domains overlapping in aggregate.

Why are domain-level overlap counts misleading? Aggregate counts pool domains across all queries, so a domain ranking for one query and AI-cited for another still counts as overlap. That inflated alignment to 165 shared domains, while per-query matching showed only about 22%. Per-query is the honest metric.

What tools do I need? A SERP API such as Apify to capture Google top 10 at scale, access to the AI engines your buyers use to collect cited domains, and a spreadsheet to store per-query lists and compute overlap.

How many queries do I need? A few dozen real buyer queries across your core categories is a reasonable start, and a couple of hundred steadies the per-query average. Treat any single run as directional, since AI answers shift by run, account, region and phrasing.