How Can You Tell Whether AI Visibility Gains Came From SEO or AEO?

Why is it hard to separate the two?

• Shared causes. Better content, clearer structure and stronger authority raise rankings and AI citations together.
• Shared pipeline. Google's AI Overviews and AI Mode draw on Google's ranking systems, and other assistants retrieve through search indexes too.
• Simultaneous work. Teams usually change content, technical settings and outreach at the same time.
• Moving platforms. Engines change how often they cite and which sources they prefer, independent of your work.
• Noisy measurement. AI answers vary from run to run, so small gains can be random.

What evidence points to SEO and what points to AEO?

No single row settles it. Several rows pointing the same way make a reasonable case.

How do you design a test that separates them?

1. Group comparable pages or questions, such as ten similar service pages or twenty questions in one topic.
2. Change one lever for one group, for example answer-first rewrites on half the pages, leaving the other half as a control.
3. Keep other work steady on both groups during the test window.
4. Record baseline and follow-up for rankings, AI citations by engine and AI referrals.
5. Compare the difference in change between the treated and control groups, not just before and after.
6. Repeat with the other lever, such as a link-building push, on a different group.

Staggering is slower than doing everything at once, but it is the most reliable way to learn which levers work for your site.

How do you diagnose a gain page by page?

For each page that gained AI citations:

• Check its rankings in Search Console for the relevant queries before and after. Search Console's generative AI performance report shows AI Overview and AI Mode impressions for the same page.
• Check your change log for what changed on that page and when.
• Check which engines cite it, using your tracking and Bing Webmaster Tools' AI Performance report for Copilot.
• Read the answers. If they quote a passage you added, the content change likely drove the citation.
• Check third-party sources cited alongside it for new mentions of your brand, which point to off-site work.

How PR shows up in AI visibility is covered in how to measure the impact of PR on AI visibility.

What role do rankings play in AI citations?

A strong one for Google's AI features, a looser one elsewhere. Ahrefs found that 38% of pages cited in AI Overviews also ranked in Google's top 10 for the same query, and many cited pages ranked lower. So ranking gains often bring AI Overview citations, but citations can also come without them. For ChatGPT, Perplexity and Claude, Google rankings are a weaker signal. In Ahrefs' 2025 study of 15,000 prompts, only about 12% of ChatGPT, Gemini and Copilot citations ranked in Google's top 10 (Perplexity 28.6%), although OpenAI reportedly used Google results for some current-events answers. So check whether the cited page's Google and Bing rankings moved before crediting a gain there to content, entity or third-party work. Gains there without Google ranking changes usually reflect content, entity or third-party work. How SEO fundamentals feed AI visibility is covered in how traditional SEO fundamentals affect AI visibility.

Which metrics should each workstream track?

Tracking leading metrics alongside outcomes makes the timing comparison possible later.

What mistakes lead to wrong attribution?

• Crediting the latest change for a gain that started before it.
• Ignoring platform-wide shifts, such as an engine citing more sources in general.
• Comparing single runs instead of repeated measurements.
• Counting mentions from branded prompts, which rise with brand awareness from any source.
• Forgetting seasonality, when demand for a topic changes over the year.
• Treating a tool's score change as proof, without checking which prompts and answers moved it.

How should you report attribution?

• Give contributions, not causes. "Rankings improved on 8 of 12 pages that gained AI Overview citations; ChatGPT citations rose on pages with no ranking change after answer-first rewrites."
• Show the control comparison. Changes in treated groups mean little without the untouched group beside them.
• Include uncertainty. Report sample sizes and intervals for AI mention rates.
• Note platform events, such as known citation drops or model changes, in the same period.
• Avoid double counting. A single gain should not be claimed by both the SEO and the AEO budget.

Connecting visibility to traffic is covered in how to correlate AI visibility metrics with actual website traffic.

What might an illustrative analysis look like?

Consider an illustrative insurance broker that runs two workstreams over a quarter: a technical and link-building programme, and answer-first rewrites of 20 guides. It rewrites 10 guides in the first month and 10 in the third, with link building spread across all pages. Guides rewritten first gain ChatGPT and Perplexity citations within six weeks while their Google rankings barely move. The second group shows no such gain until it is rewritten. Meanwhile, AI Overview citations rise across all 20 guides in step with ranking improvements from the link programme. The broker reports that rewrites drove most of the gains in ChatGPT and Perplexity. Rankings drove most of the AI Overview gains.

What do people ask about attributing AI visibility gains?

Is AEO just SEO?

Largely, for Google's AI features. For other assistants, content structure, entity clarity and third-party evidence play a bigger role.

Can tools attribute gains automatically?

Not reliably. They can show correlations, but controlled comparisons and change logs are needed to attribute.

What if we cannot stagger changes?

Use natural controls: comparable pages or questions that happened not to change, and split results by engine.

How long should an attribution test run?

At least two to three months, to allow for recrawling and the run-to-run variation in AI answers.