Answer engine optimization examples

Documented answer engine optimization examples fall into four groups: example pages that Google and Microsoft publish in their own guidance, one controlled field study of a site that rewrote its titles as questions and its summaries as short answers, studies of the pages that AI answers cite, and results companies have reported about their own AI traffic and leads. Most reported results have no comparison group, so they show what happened, not what caused it.

Every example on this page links to its source and is labeled by who published it and how it was produced. None is invented or combined from several cases. Examples that a platform wrote to illustrate its advice are labeled as illustrations; they report no results. This page is part of AEO HQ's complete guide to answer engine optimization, and the writing method the examples illustrate is set out in how to write content for answer engines. Sources were checked on September 27, 2026.

Scope and definitions

Answer engine optimization (AEO) is the work of getting a company's pages and facts found, trusted, and cited by AI assistants such as ChatGPT, Perplexity, Claude, Copilot, Gemini, and Google's AI Overviews. Generative engine optimization (GEO) names much of the same work, so the examples apply under either name.

• An example here is a documented case: a page, a change, a measurement, or a reported result, with a public source.
• An illustration is an example a platform wrote to explain its advice. It shows what the platform recommends, not what happened when someone followed it.
• A comparison group, or control, is a set of pages or a period that was not changed and is measured in the same way. It separates the effect of the work from change that would have happened anyway.
• Platform growth is the rise in AI referrals and citations that comes from the assistants' own growth. Without a comparison group, it is counted as a result of the work.

How to read an AEO example

Examples differ in what they can prove. The table sorts the ones on this page by type.

The main trap is platform growth. The authors of the only controlled field study we found write that "Public AEO success stories typically quote large raw growth multiples, but raw referral growth is confounded by the rapid platform-level growth of the answer engines themselves." (opens in a new tab) On their own site, total ChatGPT referrals grew 5.7 times while pages they had not changed grew 3.5 times (opens in a new tab) (preprint). A matched study of pages that added markup found the same pattern: AI Mode citations rose 43% from before to after the markup was added, but comparison pages gained almost as much, which left a change of 2.4% (opens in a new tab) (vendor study).

Examples from platform documentation

Microsoft's example page (illustration)

In its guidance on AI search, Microsoft Advertising shows how to write a page about a dishwasher for AI assistants. Microsoft explains that assistants such as Copilot break content into "smaller, structured pieces that can be evaluated for authority and relevance" (opens in a new tab), and gives these examples (Microsoft's illustration; it reports no test results):

Microsoft says that assistants can often lift question-and-answer pairs "word for word into AI-generated responses" (opens in a new tab). What the evidence adds: in a laboratory test of 252,000 trials across six language models, a stated price and a recent date raised the odds of being cited first in all six models, while formatting choices had no consistent effect (opens in a new tab) (peer-reviewed; the authors work for a marketing software company). Our reading: the part of Microsoft's example most likely to matter is the specific fact, 42 dB, not the list format.

Google's examples

Google's guide to generative AI search gives two examples of its own (Google's illustrations; no results reported):

• Commodity and non-commodity content. Google calls a title like "7 Tips for First-Time Homebuyers" (opens in a new tab) commodity content, "often based on common knowledge," and a title like "Why We Waived the Inspection & Saved Money: A Look Inside the Sewer Line" (opens in a new tab) non-commodity content that provides "unique expert or experienced takes that go beyond common knowledge and the ordinary" (opens in a new tab).
• One question, several searches. Google's AI features may run related searches, a technique called query fan-out. In Google's example, the question "how to fix a lawn that's full of weeds" might fan out to "best herbicides for lawns," "remove weeds without chemicals," and "how to prevent weeds in lawn" (opens in a new tab). A page can be found for one of those sub-questions.

Google also warns against the obvious response: creating separate content for every variation of a search, including fan-out queries, primarily to manipulate rankings or generative AI responses violates its scaled content abuse spam policy (opens in a new tab).

A field example: question titles and short answers

The only controlled field study of AEO page changes we found was run by two employees of Glasp on the company's own site (preprint, June 2026).

• The site. Glasp is a social web highlighter, and the pages studied were "hundreds of thousands of YouTube question-and-answer pages" (opens in a new tab) in one section of its site.
• The changes, from January 2026. Titles were converted to question form, and lead summaries were rewritten "as standalone two-to-three-sentence answers with descriptive openers." (opens in a new tab) The same bundle merged duplicate URLs into one canonical URL per video, created new pages for addresses that AI bots requested but that did not exist, locked pages that earned Google clicks against rewriting, and unpublished pages with neither search nor AI interest (opens in a new tab).
• The comparison. The rest of the same site, which was not changed, served as the control, and the authors used first-party analytics and server logs (opens in a new tab).
• The result. Against the unchanged pages, the changes were followed by an estimated 1.82-fold rise in ChatGPT referrals (95% confidence interval 1.31 to 2.54) (opens in a new tab). A stricter placebo test gave p = 0.16, so the authors call the effect "suggestive, not conclusive" (opens in a new tab).
• Search results. Google organic clicks to the changed pages did not fall beyond the site-wide trend, and the pages stayed indexed (opens in a new tab).
• Limits the authors state. One site and mostly one engine; four changes made together, so their separate effects are unknown; no random assignment; an approximate start date; and an upward trend that began before the changes (opens in a new tab).

What it shows: answer-first titles and summaries, applied with other changes to a large set of question pages, were followed by more ChatGPT referrals than similar unchanged pages on the same site, without a loss of Google clicks. What it does not show: that the same changes would work on a different kind of site, or which of the four changes mattered.

Examples of pages that AI answers cite

These studies observed which pages answers cited. They show what was cited, not whether publishing such a page would get another company cited.

Comparison pages in decision-stage answers

An agency study of about 1,000 decision-stage prompts and 57,095 citations from ChatGPT, Perplexity, Gemini, and Google's AI Overviews and AI Mode, collected January 29 to February 4, 2026 (opens in a new tab), found that comparison pages were the largest share of cited URL types, followed by "best tools" listicles, while product pages and homepages appeared at single-digit citation rates (opens in a new tab). Pages such as "HubSpot vs. Salesforce," "Monday vs. Asana," and "Shopify vs. WooCommerce" appeared repeatedly across different prompts (opens in a new tab), and third-party sources supplied about 80% to 95% of citations, depending on the industry (opens in a new tab) (agency study using a tracking vendor's data; the agency sells GEO services).

"Best" lists, including a company's own list

In ChatGPT answers to 750 prompts about software, products, and agencies, blog lists of the "best X" made up 43.8% of cited page types (opens in a new tab) (vendor study). The study gives one documented pair: Asana's own list of the best project management software placed Asana first, a similar article on Zapier's site placed Asana second, and both pages were used as sources in a ChatGPT response (opens in a new tab). Of 1,100 cited lists with clear dates, 79.1% had been published or updated in 2025 (opens in a new tab). In a separate sample, 35% of cited "best" lists were published on low-authority domains (opens in a new tab).

Agency lists in answers about agencies

An agency that tracks AI answers read 48 answers to four "which agency" prompts from ChatGPT, Google's AI Overviews and AI Mode, and Gemini, collected between August 20 and September 2, 2026 (opens in a new tab) (practitioner test; the tester kept its own agency out of its ranking). It found that the most-named agency appeared in 58% of answers, and that almost every high-frequency source was an agency's own list that ranked itself first; one agency's site was the source for 14 of the 48 answers (opens in a new tab). ChatGPT flagged the self-ranking in several answers and named the agencies anyway (opens in a new tab). Self-ranking lists carry risks, covered as entry 6 in AEO antipatterns, and how to choose an AEO agency covers what these lists mean for buyers.

Examples of changes tested with a comparison

A 2026 review of 45 studies found that no reviewed technique shows a stable, longitudinal, cross-platform causal effect on organic discoverability (opens in a new tab) (preprint). The full record of which findings hold up is in what replicates in AEO and GEO research.

Examples of results companies reported

The table records five public reports of AI traffic, leads, or growth: what each states, and what it leaves out.

Three points follow. First, only the Glasp study above separated the work from platform growth; against that study's 3.5-fold growth in untreated pages, a raw multiple says little on its own. Second, the Optimist case page states the limit itself: "directly attributing any specific action to any specific result isn't something we — or anyone — can do with certainty." (opens in a new tab) Third, the Ahrefs and Omniscient figures are measurements, not results of AEO work: they show how much of each company's traffic or leads came through AI, which is the starting point for the metrics in AEO metrics and KPIs.

What the examples show and do not show

Antipatterns

An antipattern is a practice that looks helpful but fails or backfires. Six come up when companies copy AEO examples:

1. Copying the format without the facts. In the laboratory, a price and a recent date moved citations, while formatting did not (opens in a new tab). Instead, copy the specific, checkable facts first.
2. Quoting a raw growth multiple. Untreated pages grew 3.5 times (opens in a new tab) in the one controlled study. Instead, compare changed pages with unchanged ones.
3. Publishing a list that ranks yourself first because lists get cited. ChatGPT flagged self-ranking (opens in a new tab) in a published test. Instead, publish comparisons with a stated method, and leave yourself out of any ranking.
4. Making a page for every sub-question. Google treats pages for every variation, including fan-out queries, made primarily to manipulate AI responses as scaled content abuse (opens in a new tab). Instead, answer close variants on one page.
5. Treating a vendor's customer result as evidence for your site. HubSpot's comparison of users and non-users (opens in a new tab) does not say how the groups were matched. Instead, look for a comparison group and a stated method.
6. Rewriting pages that already earn search clicks. The field study locked pages with meaningful Google clicks against rewriting (opens in a new tab), and its Google clicks held. Instead, protect pages that earn search traffic when you change others.

Checklist: judging an example before you copy it

FAQ

What is an example of answer engine optimization?

The best-documented example is Glasp's: it rewrote titles as questions and opening summaries as short, standalone answers (opens in a new tab) on a large set of question pages, among other changes, and measured ChatGPT referrals against pages it left unchanged. The result was an estimated 1.82-fold rise, which its authors call suggestive, not conclusive.

What does an AEO-optimized article look like?

It answers one question near the top, in the reader's words, and then gives specific facts, such as Microsoft's "It operates at 42 dB" (opens in a new tab). How to write content for answer engines sets out the steps and the evidence for each.

Are there AEO case studies with real results?

A few publish their method: the Glasp field study, the matched schema study, and Omniscient Digital's attribution analysis above. Many published case studies report growth multiples without a baseline or comparison group, so they cannot separate the work from the assistants' own growth.

Which brands are winning with AEO?

No study shows which brands' AEO work caused their visibility. Studies do show who appears: in one panel, first answers named household brands 73% of the time, mid-market brands 44%, and small brands 11% (opens in a new tab) (preprint; the author co-founded the platform studied). Well-known brands start ahead of any optimization. To see how your own company appears, use the method in how to measure AI visibility.

How do AEO results compare with SEO results in these examples?

In the field study, Google organic clicks to the changed pages did not fall beyond the site-wide trend (opens in a new tab) while ChatGPT referrals rose relative to unchanged pages. It is one site, so it does not show that AEO changes are always safe for search.

Are AEO examples different from GEO examples?

Mostly not, because the terms describe much of the same work. Google says AEO and GEO "are both terms you may see used to describe work specifically focused on improving visibility in AI search experiences" (opens in a new tab). AEO vs GEO compares the terms.

How long did the documented changes take to work?

We found no study that measures how long a change takes to show up in AI answers. The field study found an increase aligned with changes made in January 2026, but its authors could only approximate when the rollout began (opens in a new tab).

How do I track AI referral traffic from changes like these?

Set up a channel for AI assistants in your analytics, as described in how to track AI referral traffic in GA4, and compare the pages you change with pages you leave alone.

Next steps

AEO HQ sells AEO services at fixed, published prices, listed on the pricing page. AEO HQ's own work does not appear as an example on this page, because we have no published result that meets the checklist above.

Change log

• September 28, 2026: First published.