Say you run a site with forty articles, a few hundred search clicks a month and about ten hours a week for it. Should those hours go into SEO or into AEO (getting cited in AI answers)?
For most small sites the order is SEO first, then a thin AEO layer on the pages that already rank. AI engines pick their sources mostly from pages they can retrieve near the top of a search, so a page nobody can find won't be cited however well it's written. Once a page ranks, the AEO work on it is cheap and often decides whether you keep the click. If the terms themselves are new to you, what is AEO defines them; this piece is about order of work.
Why ranking comes first
Three findings point the same way.
Retrieval position beat every content tactic in the lab. C-SEO Bench (NeurIPS 2025) tested ten content optimization methods on four language models. In its retail task on GPT-4o-mini, placing a document first in the list the model read improved its citation rank about 7.7 times more than the best content method. On Claude 3.5 Haiku the best content method was slightly negative. Does GEO work? has the full comparison.
ChatGPT mostly cites pages that rank. AirOps analyzed 548,534 pages ChatGPT retrieved for 15,000 prompts: 55.8% of cited pages ranked in Google's top 20 for the prompt or a related fan-out query, and pages ranking first were cited 43.2% of the time, about 3.5 times as often as pages outside the top 20.
But not only page one. Ahrefs found 37.9% of AI Overview citations came from pages in Google's top 10, down from 76% seven months earlier (Ahrefs says better detection on its side explains part of the drop). Pages ranking 11 to 30 can still be cited, so AEO work on them isn't wasted.
Put together: rankings get you into the candidate pool, and AEO work decides how often you're picked from it.
Sort your pages before you choose
The answer to "SEO or AEO?" differs page by page. Search Console gives you enough to sort them in an afternoon.
- Check that every engine can reach you. robots.txt, your CDN's bot settings, and the Search Console Search generative AI control (worldwide since August 31, 2026; if it's switched on, you're out of AI Overviews and AI Mode). The free crawler checker covers the robots.txt part.
- Export page-level data from the Performance report for the last 3 months against the same period a year earlier: clicks, impressions, CTR, average position.
- Put each page in one of four buckets:
| What Search Console shows | What it usually means | Do first |
|---|---|---|
| Not indexed, or average position worse than ~30 | The engines can't find it either | SEO: indexing, internal links, a better match for the query |
| Position 11–30, impressions growing | In the candidate pool, rarely picked | Both: improve the page for ranking, and add the AEO layer |
| Position steady in the top 10, CTR and clicks falling | An answer on the results page is absorbing the visit | AEO: make it the source the answer cites |
| Position and CTR steady | Working | Leave it; refresh facts on a schedule |
- Cross-check the third bucket in Search Console's AI performance report, which shows impressions in AI Overviews and AI Mode by page (no clicks, no queries). Pages with rising AI impressions and falling clicks are your first AEO candidates.
The third bucket is where AEO pays fastest, and it's worth it: Seer Interactive's 2026 data shows brands cited in an AI Overview earning about 120% more organic clicks per impression than uncited ones on the same kind of query. The zero-click article has the rest of the click data.
What the AEO layer actually involves
On a page that already ranks, it's a short list, and most of it also helps the page in classic search:
- The first 50 to 80 words answer the page's main question with no preamble; each section opens with its own answer
- Sections that make sense on their own, roughly 40 to 120 words, so an engine can lift one without its neighbors
- Every statistic linked to a named source
- A comparison table where the page compares things
- A named author with a bio
- Facts checked and the visible content updated, not just the date
What it doesn't involve: new markup formats or "AI files". Google says there are no additional requirements for AI Overviews or AI Mode. Schema that you'd add for classic rich results is still worth adding; it just isn't a separate AEO lever.
Where SEO still deserves most of the time
- Transactional queries ("buy", "pricing", "coupon"), where AI answers appear less often and the user wants to act
- Interactive pages (calculators, tools), which an answer can't replace
- Display-ad sites, where a citation without a visit earns nothing
- Any site whose technical basics are shaky: crawl errors, thin internal linking, slow templates
Where AEO earns its hours sooner
- Comparison and "best X" queries, which AI answers handle by default
- Definitions and explainers, where people increasingly ask an assistant instead of reading ten results
- Pages already in the top 10–20 losing clicks, per the table above
A 10-hour week
One workable split for a small site with a sound technical base. It's my own rule of thumb, not a measured optimum:
- 5 hours: SEO. One substantial new page, on-page fixes, internal links, technical hygiene
- 2 hours: the AEO layer on existing pages from the third bucket: openings, self-contained sections, sources
- 2 hours: distribution and mentions. Getting talked about in your niche. Ahrefs found web mentions correlate more strongly than backlinks with brand visibility in AI Overviews (0.664 against 0.218, 2025)
- 1 hour: measurement. Search Console, plus a citation sample (below)
Shift time toward AEO when rankings are healthy and clicks are slipping. Shift it back toward SEO when pages aren't being indexed or ranked at all.
Measuring each
SEO is well served: Search Console for Google performance, a keyword tool for rankings and links, analytics for conversions.
AEO measurement is sampled. Search Console now reports AI Overview and AI Mode impressions per page, and GA4 added an AI Assistant default channel in May 2026 for visits that arrive with an AI referrer. Neither tells you which prompts you were cited for. For that you either use a tracker or sample by hand:
- Pick 10 to 20 prompts that match pages you care about.
- Run each several times over a few days in each engine your audience uses.
- Record whether your domain is cited, and track the rate over weeks rather than any single answer.
Choose search-style prompts for the sample. Most ChatGPT conversations never trigger a search: a Similarweb estimate for May 2026 found only 6.8% of all US desktop ChatGPT prompts produced a citation, so random prompts would mostly measure nothing. Perplexity, which cites on nearly every answer, gives you a readable rate fastest.
FAQ
Can AEO changes hurt my Google rankings?
Do I need an llms.txt file for AEO?
My site is brand new. Should I bother with AEO at all?
Sources
- Puerto et al. — C-SEO Bench: Does Conversational SEO Work?, NeurIPS 2025 (arXiv 2506.11097)
- AirOps — The influence of retrieval, fan-out and Google SERPs on ChatGPT citations, March 2026
- Ahrefs — AI Overview citations from the top 10, 2026
- Seer Interactive — AIO impact on Google CTR: 2026 update, April 2026
- Ahrefs — AI Overview brand correlation study, May 2025
- Google — AI features and your website
- Google Search Console Help — Search generative AI control
- Search Engine Roundtable — GA4 AI Assistant channel, May 2026
- Relevant Audience — Similarweb ChatGPT citation rates by sector (May 2026 US desktop data)
Updated October 1, 2026: rewritten as a practical order-of-work guide for small sites; updated the click data and measurement options, and corrected the ChatGPT citation-rate source.