There is a lot of noise about answer engine optimisation and most of it is search advice from 2015 with a new label. Here is what actually differs, from someone who spent a day doing it and measuring the result.
The core difference in one sentence
A search engine ranks pages. An answer engine extracts sentences. Those are different jobs, and a page optimised for the first can be useless for the second.
Ranking rewards a page that is comprehensively about a topic. Extraction rewards a page containing a self-contained sentence that answers a question without needing the paragraph around it. You can be excellent at one and invisible in the other.
Write sentences that survive being removed
The test: cut any sentence out of your page, show it to someone who has not read the rest, and see whether it still means anything.
Does not survive: "We founded it in 2025 and have been growing ever since." Who is we. Founded what.
Survives: "Purcell Ventures LLC is a Georgia software company founded by Elijah Purcell in April 2025 that builds websites and software for small businesses."
The second is worse prose and better extraction, and the fix is not to write the whole page like that. It is to make sure the load-bearing facts appear once each in that shape, usually near the top, and then write normally.
Third person for facts about a person
A page that says "I am nineteen and I run a software company" cannot be lifted and presented as a fact about a named entity, because a model has no way to attach "I" to anything.
This is why I have two pages about the same person. An About page in first person, which is how a human wants to read it, and an answer page in third person, which is the shape an engine can use. They link to each other and say which is which.
FAQPage schema, and the discipline it needs
FAQPage is the format that hands a machine a question and a self-contained answer as structured data.
Correction, added the same day this was published. An earlier version of this paragraph called FAQPage the single highest-leverage item here. That was true when it produced a rich result in Google, and it is not true now: Google stopped showing FAQ rich results on 7 May 2026 and is removing the reporting and testing support this summer. The schema type is still valid, and it is still read by Bing and by the retrieval crawlers behind the assistants, which is what this article is about. So it keeps its value for answer engines and has none left for Google results, and saying otherwise would be selling you something that stopped existing three months ago.
The discipline: every answer must be genuinely answered in the visible body of the page. Schema promising an answer the page does not contain is the structured data equivalent of a misleading headline, and it is treated as one.
Write the answers as though they will be read alone, because they will be. No "as mentioned above." No "our." Name the thing every time.
One identity, stated the same way everywhere
If your site names a person or company in several places, every one of those structured data nodes needs the same @id. Otherwise you are describing several entities that happen to share a name, on your own domain.
Mine was twelve distinct Person nodes, three of them anonymous, before I checked. It should have been one, and now is.
The same applies to prose. Pick one bio sentence and use it identically on every profile. Three slightly better sentences on three platforms read as three people who might be the same person. One adequate sentence repeated exactly reads as one.
Say the disambiguation out loud
If somebody shares your name, do not hope. schema.org has disambiguatingDescription for exactly this, and the plain-prose version belongs on the page too.
Three other people share mine. My answer page names them and says which one I am not, because a sentence is cheaper than letting an assistant hedge.
llms.txt, briefly and honestly
A plain-text file at the root of your site describing what is on it and where. Adoption is not universal and nobody should promise you results from it.
It costs twenty minutes and it forces a genuinely useful exercise: writing down what each of your pages is actually for, in a sentence, without marketing. Half the value is that most sites cannot do it.
The one with the largest measured effect
Cite external sources inline, and link them.
This is the technique with the strongest reported result in the generative engine optimisation literature, and the effect is largest for sites that are not already ranking well: a 2026 survey of the field reports substantial citation gains for pages that cite authoritative sources inline, with the biggest lifts going to lower-ranked sites. The same body of work finds statistics, definitions, prices, dates and quotations improve citation by roughly a quarter, and that keyword stuffing measurably reduces it.
The uncomfortable corollary, and the reason I am putting it here rather than burying it: that survey found no established evidence that content volume, publishing velocity or structured data are primary optimisation vectors. Most of what is sold as answer engine optimisation is those three things.
What matters more than all of it
Being indexed at all.
I did every technique above and then checked how many of my pages Google actually had. Seven, out of twenty-eight submitted. Every clever piece of structured data was sitting on pages nothing could see, and the cause turned out to be one line of inherited metadata telling Google to skip them.
So before optimising for extraction, verify you are in the index. The nine checks are here. Structured data on an unindexed page is a very tidy way of achieving nothing.
Common questions
What is answer engine optimisation?
Optimising a page so that an AI assistant can extract a correct answer from it, as distinct from ranking the page in a list of results. A search engine ranks pages; an answer engine extracts sentences. A page can be good at the first and useless for the second.
How do I get an AI to quote my website?
Write the load-bearing facts as self-contained sentences that still mean something when removed from the page. Use third person for facts about a person, since a model cannot attach an I to anything. Add FAQPage structured data where every answer is genuinely answered in the visible body. Use one identifier for your entity across every structured data node, and one identical bio sentence across every profile.
Does llms.txt actually work?
Adoption is not universal and nobody should promise results from it. It costs about twenty minutes and it forces a useful exercise, which is writing down what each page on your site is actually for in one sentence without marketing language. Most sites cannot do that, and finding out is worth the time on its own.
What is the most common reason AI does not mention my site?
Usually that the pages are not indexed rather than that they are badly written. Verify how many of your pages are actually in the index before optimising anything about their content. Structured data on an unindexed page achieves nothing.
Should I write a separate page in third person about myself?
If you want facts about you to be quotable, yes. A first-person About page is how a human wants to read it, and a third-person answer page is the shape an engine can lift and attach to a named entity. Keep both, link them to each other, and make clear which is which.