Why AEO Best Practices Might Not Be Enough for AI Citations

What changed in the evidence for AEO best practices?

The evidence changed on November 10, 2026, when Christopher Jan Benitez tested whether pages following more common AEO practices were more likely to be cited. Across 51 US queries and three engines, cited pages did not consistently outscore pages the same engine read or ranked but skipped. That does not establish that clear answers, crawlable pages or sources are useless. It establishes that a checklist alone did not separate cited from skipped pages in this sample.

The study scored pages on 15 checks: accessibility to bots, answer-first structure, short paragraphs, headings, lists or tables, direct statements, dates, named entities, source links, author information and valid structured data. Google AI Overviews cited pages with an average score of 64.9, while skipped top-10 pages averaged 68.0. ChatGPT cited pages averaged 64.8 against 69.0 for pages it read but did not cite. Perplexity was the one engine where cited pages scored higher, 65.6 against 63.3, but its reported range included the no-signal midpoint.

The useful shift is operational. Teams should stop treating an AEO score as a proxy for citation performance. Keep the elements that make content readable, indexable and auditable. Then test whether a page earns citations for the questions that matter to the category. Citation Share measures that result directly: the percentage of relevant AI answers in a category that cite your brand.

Google makes a related distinction in its own documentation. A page must be indexed and eligible to show a snippet to be eligible as a supporting link in AI Overviews or AI Mode, but Google says there are no additional technical requirements or special optimizations for those features. Eligibility is a floor. Citation selection is the outcome that needs monitoring.

1. Before November 10, 2026: common AEO advice was often treated as a likely route to citations.
2. After the study: cited pages scored higher than skipped pages only 44% of the time in Google AI Overviews, 41% in ChatGPT and 55% in Perplexity.
3. What to do now: retain technical and editorial hygiene, then run prompt-level citation measurement across the engines your buyers use.
4. What to avoid: publishing more checklist-shaped pages without checking whether they are cited.

Why did the checklist fail to predict citations consistently?

The checklist failed to predict citations consistently because engines appeared to favor different source types before individual page-format signals. In the study, Google AI Overviews cited YouTube for 36% of sources. ChatGPT used brand homepages for 27% of its sources, often as attached links to recommended brands. Perplexity cited longer articles more often. Those source-type patterns were larger than the differences associated with the 15 AEO checks.

This matters because many AEO rules were designed around article pages. A homepage can be the right destination for a brand recommendation while scoring poorly on article-focused checks such as question headings and answer capsules. A YouTube video can be selected even though a publisher cannot apply the same page structure. A long explanatory page can suit one engine without becoming a universal citation format.

The study also identified a measurement limit that should shape interpretation. Google does not disclose the pages an AI Overview read, so the comparison used top-10 organic pages that the overview skipped. ChatGPT and Perplexity exposed pages read during the search process, which enabled a closer cited-versus-read-and-skipped comparison. The data was collected on one day in the United States, and Perplexity was tested through its Sonar API rather than its consumer app.

That is enough evidence to challenge certainty, not enough to declare a permanent rule for every market or model. AI answers vary by query, date, location and engine. Content teams need a repeatable observation layer that records the answer, cited domains, cited URLs and competing brands over time. A monthly Answer Presence view shows whether a brand appears across the relevant question universe, while Citation Share shows how often it earns the supporting link.

1. Google AI Overviews: source selection included a substantial YouTube share in this sample.
2. ChatGPT: brand homepages appeared far more often among citations than among pages read and skipped.
3. Perplexity: longer articles were more common among citations in the reported sample.
4. All engines: the observed source pattern was more informative than a universal AEO checklist.

Who does this affect most?

This affects B2B SaaS and technology growth teams first, because their core queries often ask AI to recommend, compare or explain products. A buyer asking for the best category tool may receive a mix of review pages, documentation, comparison content and links to official homepages. A content program that only ships answer capsules can miss the source format that a specific engine selects for that prompt.

Local and multi-location service businesses face the same problem through different questions. Queries such as best service in a city can draw on business pages, directories, reviews, local sources or videos. The operational question is whether the business appears and is cited for the service and geography combinations that lead to calls or bookings. Broad organic traffic alone cannot answer that question.

Editorial teams are also affected. A strong, sourced article remains worth publishing because it can satisfy a reader, support search visibility and provide material an AI system can cite. Yet content production needs broader coverage than a single ideal template. Build useful pages around the actual information needs in a cluster: definitions, comparisons, alternatives, implementation questions, first-party proof and current updates.

Agencies and in-house SEO teams should adjust reporting as well. A completion report that says pages meet AEO best practices does not prove citation performance. Report the pages published, their technical checks, the prompts monitored, the engines tested and the resulting Citation Share. That creates a clean separation between work completed and the outcome the work is intended to influence.

1. B2B SaaS teams should monitor category, alternative and comparison prompts.
2. Service businesses should monitor service-plus-location prompts and AI Overview presence.
3. Editorial teams should match source formats to question intent instead of relying on one page template.
4. SEO leaders should report citation outcomes alongside crawlability, indexing and content production.

How should content teams respond without abandoning AEO?

Content teams should keep AEO as baseline publishing discipline, then make citation measurement the decision system. Google states that standard SEO practices remain relevant for its AI features, including meeting technical requirements and creating helpful, reliable, people-first content. That supports preserving fast, accessible pages with clear information and meaningful internal links. It does not support assuming a special markup or an answer-first paragraph guarantees an AI Overview citation.

Start with an engine and prompt inventory. List the commercial, informational, comparison and local questions that a buyer could ask before choosing a provider. Run the same prompt set across ChatGPT, Perplexity, Gemini, Copilot and Google AI Overviews where applicable. Record whether the brand appears, whether it is cited, which URL is cited and which competitor domains are selected.

Then inspect the pages that repeatedly earn citations. Identify their source type, the claim they support, their freshness signals, their primary evidence and the page format. The goal is not to copy a surface has from one winner. It is to find gaps in coverage that the current content program can fill with stronger evidence. A cited source that provides a dated comparison, a clear methodology or primary documentation gives an engine something concrete to use.

Finally, turn the findings into publishing decisions. If category prompts cite comparison pages, publish comparison pages backed by verifiable evidence. If brand recommendations attach official sites, make important product information easy to find on official pages. If local questions cite location-specific sources, create complete location pages that answer the actual service question and reflect current operations. Omnicite calls this Citation Engineering because the work joins coverage, authority, freshness and measurement.

1. Keep crawlability, indexing eligibility and people-first writing as baseline controls.
2. Measure the same prompt set across relevant AI engines on a recurring schedule.
3. Classify cited URLs by source type, intent, evidence and freshness.
4. Prioritize the coverage gaps that can raise Citation Share for high-intent questions.

What should a citation measurement loop include?

A citation measurement loop should include a fixed prompt universe, an engine-level capture process, a source classification method and a review cadence. Without a fixed prompt universe, a rise or fall in citations can reflect a changed question set rather than changed visibility. Without URL-level capture, a team can see a brand mentioned but miss which page an engine trusted enough to link.

Use four outcome measures. Citation Share is the headline percentage of relevant answers that cite the brand. Citation Count per day measures volume. Answer Presence measures how broadly the brand appears across the tracked questions. Share of Voice compares the brand with named competitors. These measures answer different questions, so do not combine them into one result without retaining the underlying evidence.

The study suggests another practical check: record source type. A citation to a homepage, editorial article, video, government page, review or directory signals different content and distribution needs. It also prevents a weak diagnosis. If an engine cites a competitor homepage for recommendation prompts, publishing another long article may leave the actual gap untouched.

Use recurring reviews to identify change, not just snapshots. The cited sources in AI answers can shift with query wording, source freshness and engine updates. Compare each run with the previous run using the same prompts, country and engine settings. Mark any methodology change in the log. That allows a team to see whether a new page increased Citation Share or whether the measured environment changed first.

1. Prompt ID and exact wording, including market and language where relevant.
2. Engine, date and captured answer with cited domain and URL.
3. Brand presence, citation status and competitor citations for each prompt.
4. Source type, content gap, owner and next publishing or refresh decision.

What should you publish next?

Publish the page that closes a measured citation gap for a high-intent question. The next page may be a category comparison, a buyer guide, a current definition, a location page or an official product explainer. Its format should follow the evidence from the prompt set, not a generic AEO template.

For every page, make the claim easy to verify. Put the answer near the top when that helps the reader. Use dated primary sources for numerical claims. Explain scope and limitations. Maintain the page when facts change. Those practices produce better editorial work and reduce technical barriers, even if they cannot guarantee selection by a particular engine.

The November study has a useful constraint for teams chasing a shortcut. No single tested AEO practice made a consistent difference across Google AI Overviews, ChatGPT and Perplexity. The right response is broader measurement, stronger source material and coverage that matches the question set. Citation work earns confidence through observed results, not a checklist score.

Measure the next publishing cycle against the same prompts. Keep pages that earn citations current, investigate pages that are present but uncited and build the missing evidence where competitors keep winning. That is a concrete route from AEO hygiene to a citation program that can be evaluated.

1. Choose one high-intent prompt cluster with a visible citation gap.
2. Publish or refresh the page format that the measured engine selection makes credible.
3. Cite primary sources and state dates, scope and limits clearly.
4. Re-run the identical prompt set and compare Citation Share after the page is discoverable.

Sources

Source: Christopher Jan Benitez

Across 51 US queries and three engines, cited pages did not consistently score higher on 15 AEO best-practice checks; the article reports engine-specific comparisons and study limits. Christopher Jan Benitez, 2026-11-11

Source: Google Search Central

Google states that no additional requirements or special optimizations are needed for AI Overviews or AI Mode, while pages must meet Search eligibility requirements to appear as supporting links. Google Search Central, 2025-12-10