Search is splitting in two. One half still looks like ten blue links. The other half is a paragraph written by a model that read your page, decided whether to trust it, and either named you or did not.
Answer engine optimization is the work of winning that second half. It is not a rebrand of SEO — the foundations are shared — but the optimisation target is genuinely different. You are no longer competing for a position in a ranked list. You are competing to be the passage that gets lifted.
What is answer engine optimization?#
Answer engine optimization is the practice of structuring content so that an AI assistant can extract a specific passage, verify that it is correct, and attribute it to you. Three verbs, three separate failure modes.
Most pages fail at extraction. The answer exists, but it is spread across four paragraphs of preamble, and the model cannot lift a clean self-contained chunk. Some fail at verification: the claim is there, stated confidently, with nothing behind it. A few fail only at attribution, which is usually a schema problem rather than a content problem.
Why the usual SEO checklist is not enough#
A conventional checklist optimises for a crawler building an index. An answer engine is doing something closer to reading comprehension under time pressure. It chunks your page, scores each chunk for relevance, and keeps the ones it can defend.
That changes what matters:
- Keyword density stops mattering almost entirely. Semantic coverage of the topic replaces it.
- Passage boundaries start mattering a lot. A chunk that needs the previous three paragraphs for context will be dropped.
- Sourcing becomes a ranking factor, not a nicety. A model choosing between two similar passages will prefer the one that cites primary sources.
The six levers that actually move AI citation#
1. Lead with the answer, in the first 60 words#
The most valuable real estate on the page is the block immediately after the headline. Put a direct, complete answer there — 40 to 70 words, no preamble, no throat-clearing.
This feels wrong to write. It gives away the payoff before the reader has invested anything. Do it anyway. Assistants quote this block far more often than any other part of the page, and readers who want the detail keep scrolling regardless.
2. Shape headings like the questions people actually ask#
"Implementation considerations" is invisible to an answer engine. "How long does answer engine optimization take?" matches a real query string.
One or two question-shaped H2s per article is enough. They also earn People Also Ask placements, which is a useful side effect.
3. Make every FAQ answer stand alone#
An FAQ answer gets quoted without its question attached surprisingly often. If the answer starts with "It depends on your setup", the quote is worthless and the model learns to skip you.
Write each answer as though it were the only sentence anyone will ever read from the page.
4. Say what things are, in schema#
FAQPage marks a block as question-and-answer pairs. Article with a real author node and datePublished tells the model who is talking and when. about and mentions disambiguate which entities the piece concerns.
None of this is exotic. Almost nobody does it properly, which is exactly why it works.
5. Cite your sources, outbound#
A claim with a link to a primary source outranks the same claim without one. This is the cheapest trust signal available and the one marketers resist most, because it means sending a reader elsewhere.
Send them. The citation is what makes you quotable.
6. Publish a clean text twin of every page#
HTML is lossy to parse. Navigation, cookie banners and layout markup all add noise the model has to discard.
Publishing a Markdown version of each article — plus an llms.txt index at the site root — removes that loss entirely. Every article on this site is available at /blog/<slug>/index.md for exactly this reason.
What to measure#
Traditional rank tracking will not tell you whether this is working. Track instead:
| Signal | How to check it | Healthy direction |
|---|---|---|
| Direct citations | Ask the assistants your target questions monthly | Named more often |
| AI crawler hits | Server logs, filtered by user agent | Rising |
| Referral traffic from assistants | Analytics, chat.openai.com and similar | Rising |
| Branded search volume | Search Console | Rising |
The last one is the leading indicator most teams miss. Being cited in an answer often produces no click at all — it produces someone searching your brand name a week later.
Where to start#
If you do one thing this week, add an answer-first block to your ten highest-intent pages. It takes an afternoon and it is the change with the shortest path to a citation.
If you want the rest of it run properly — schema, sourcing, the text twin, the measurement loop — that is the SEO and AEO work we do for clients.
