SEO gets you found. AEO gets you quoted.
Search engine optimisation earns a position on a results page. Answer engine optimisation is the practice of structuring a site so an AI assistant can quote it accurately when somebody asks a question it answers. They share most of the same work, and the part where they differ is worth naming separately.
Personal reply from Andrew within one business day.
What is answer engine optimisation?
Answer engine optimisation is the practice of structuring a site, and the information about a business elsewhere on the web, so that an AI assistant can quote it accurately and attribute it correctly when someone asks a question the site answers.
It is not a replacement for search engine optimisation, and treating it as one is the first mistake available. It is what you do once you accept that a growing share of buyers will never see a results page at all — they will ask, receive a paragraph, and act on it. In that exchange there is no list of ten blue links to be eighth on. There is one answer, and either the business is in it or it is not.
The mechanism is different from ranking, and that difference drives the work. A search engine ranks documents against a query. An assistant assembles a claim and then looks for grounds to make it confidently — which means it favours statements that are specific, self-contained, attributable to a named entity, and consistent with what it can find elsewhere. A page can rank respectably and still never be quoted, because it hedges, because the answer is buried, or because the business is described three different ways across three profiles and nothing corroborates anything.
Most of AEO is the unglamorous half. Naming the business the same way everywhere. Publishing a description that does not drift between the site, the register entry and every directory. Marking up who the organisation is and what it offers so the relationship is stated rather than inferred. Answering real questions in short, quotable sentences instead of in three paragraphs of throat-clearing. None of it is exotic, and doing it consistently is genuinely rare, which is why it still moves the needle.
The other half is honesty about what cannot be done. Nobody controls what an assistant says. There is no submission endpoint, no ranking factor to buy, and no way to make a model repeat a flattering claim that nothing outside your own site supports. What can be done is to make the accurate answer the easiest one to assemble — and to find out what is currently being said, which is usually the least comfortable slide in any presentation on this subject.
The homepage carries three worked illustrations of what a correct answer would look like for this business, with the reasoning attached. They are illustrations rather than transcripts, which is the honest way to show the target without implying a result nobody can promise.
How does a question become a citation?
Five stages, each of which can fail on its own. The work is a matter of finding which one is failing rather than doing all five at equal intensity.
-
01
The question
A buyer asks in their own words — typed into a search box, or increasingly spoken at an assistant that will answer rather than list.
breaks here if: nothing on the site is phrased the way anybody actually asks -
02
The page
One page owns that question and answers it completely in its first paragraph, without needing the rest of the page to make sense.
breaks here if: the answer is spread thinly across four pages and complete on none -
03
The structure
Schema names who the business is, what it offers and what it just answered — and the markup says the same thing the visible page says.
breaks here if: the markup and the page disagree, which is worse than no markup -
04
The corroboration
The same name, the same description and the same contact details wherever else the business appears on the web.
breaks here if: three profiles describe what reads as three different businesses -
05
The citation
The assistant has something specific, attributable and consistent to repeat, and repeats it.
breaks here if: a competitor is simply clearer about the same claim
Stage two is where most sites fail, and it is the least expected. A business will have a page that genuinely covers the topic, written by someone who knows it, and still not be quoted — because the answer arrives in the fourth paragraph after the context, or because it is split across a services page and a blog post and neither is complete alone. An assistant summarising that page takes what is at the top. If what is at the top is scene-setting, the scene-setting is what gets attributed to you.
Stage four is where most of the effort goes, and it is almost entirely off the site. This is the part that cannot be fixed by writing better pages: if the register entry, the professional profiles and the directory listings describe the business in three different ways, there is nothing for a model to corroborate against and it will hedge or reach for a competitor who is easier to be confident about. Fixing it is tedious, one profile at a time, and it is frequently the single highest-value thing available.
Stage three is the one people try to buy first, because schema markup feels like the technical lever. It matters, and it is not a shortcut: markup that asserts something the visible page does not say is worse than no markup at all, both because search engines treat the mismatch as a quality signal and because it is precisely the kind of inconsistency stage four is about. On this site every structured-data block is generated from one source of truth and the FAQ answers are extracted from the visible page, so the two cannot drift apart. That is not perfectionism — it is the cheapest way to stop the failure mode.
How do SEO and AEO actually differ?
Enough to be worth separating, not enough to be sold as two unrelated engagements. Roughly two-thirds of the work serves both.
| SEO | AEO | |
|---|---|---|
| What it competes for | A position on a list of results, against everyone else targeting that query. | Being the sentence an assistant lifts. There is usually room for one or two, not ten. |
| What a win looks like | Page one for a term that buyers actually type, and a click that follows. | Being named accurately, with the right specialty attached, in an answer nobody clicks through from. |
| Where the content effort goes | Depth and coverage. Longer pages, more of them, internally linked. | Precision and self-containment. Short definitional sentences that survive being extracted. |
| What decides it off-site | Links from places that already have authority. | Consistency. The same entity described identically in every place it appears. |
| How fast it moves | Months. Slow, reasonably predictable, and reversible if the work stops. | Unpredictable. Model updates can change an answer overnight in either direction. |
| How it is measured | Rankings and traffic, from tools built for the purpose over twenty years. | Asking a fixed set of questions on a fixed schedule and recording what comes back. |
| The failure mode | Ranking for terms that bring visitors who were never going to buy. | Being described confidently and wrongly, which is worse than not being mentioned. |
The last row is the one worth sitting with. A business that is not mentioned by an assistant has a marketing problem. A business that is mentioned with the wrong specialty, the wrong location, or a service it stopped offering three years ago has a correction problem, and corrections propagate slowly. That is the argument for finding out what is currently being said before deciding what to change — and it is why the first deliverable is a baseline rather than a plan.
What does the work actually involve?
Three phases, in a fixed order, because each one is wasted if the one before it has not been done.
Find out what is true now
- Crawl, index and redirect audit
- Core Web Vitals and page speed
- Which page currently owns which term
- What assistants say today, recorded
Nothing is decided in this phase. It exists so the later work is aimed at what is actually failing rather than at what is easiest to sell, and so there is a number to compare against later.
Make the site answerable
- Keyword-to-page map, one owner per term
- Answer-first rewrites where the answer is buried
- Organisation, Service and FAQ structured data
- Entity naming reconciled everywhere it appears
This is where the structural work happens, and it overlaps heavily with a rebuild. If the page set is wrong, this phase becomes a rebuild whether or not it was scoped as one.
Answer questions people ask
- Question research from actual enquiries
- A page per question worth owning
- Internal linking that connects the claims
- A cadence the business can sustain
The slow half, and the one most engagements abandon. A sustainable monthly cadence beats an ambitious weekly one that stops in March, because the compounding is the entire mechanism.
- Terms ranking on page onetracked monthly
- Assistant citationsfixed question set
- Enquiries per monthsource attributed
- Audit requests bookedthe one that matters
Measured monthly, against a question set fixed at the start of the engagement and not quietly edited afterwards. A question set that moves can make any trend look like progress, so the list is agreed once and then left alone. No figures are published on this page, because a target quoted before a baseline exists is a guess with a decimal point on it.
Want to know what an assistant currently says about your business? That is the first thing the audit checks, and it is free.
Book a free process auditWhen is this not worth doing?
This is a field with a long history of confident promises, so the limits are worth stating plainly. Four of them decide whether the work makes sense at all.
There is nothing to optimise yet. If the site has four pages, none of which owns a term or answers a question completely, this work becomes a rebuild in everything but name. That is fine, but it should be scoped and priced as one rather than discovered halfway through an SEO retainer.
Your customers are a known list. A business that sells to twelve accounts it already has, or wins work entirely through a tender process or a professional network, is not held back by search visibility. Money spent here buys very little, and the automation and integration pillars are almost certainly where the return is.
Nobody can answer more enquiries. Presence work increases the number of conversations. If the current ones are already being answered late, more of them makes the business worse rather than better, and the honest sequence is to fix stage four before widening stage one. The services overview sets out that chain in order.
Nothing here is guaranteed, by anyone. No agency controls search rankings and none controls what an assistant says — there is no submission endpoint and nothing to buy. A guaranteed position is either a term nobody searches for or a claim that will not survive contact with a model update. What can be committed to is the work, the baseline, and the monthly measurement against a question set that does not move.
Questions about SEO and AEO.
What is the difference between SEO and AEO?
SEO earns a position on a page of search results, which somebody then clicks. AEO earns a mention inside an answer an AI assistant gives directly, which usually nobody clicks. They share the same foundation — a crawlable, fast site where each page answers one question completely — and diverge at the end: SEO rewards depth and inbound links, while AEO rewards short self-contained statements, structured data that matches the visible page, and a business described identically everywhere it appears.
Can you guarantee a position on page one, or a mention by ChatGPT?
No, and neither can anyone else. Search engines do not sell ranking positions and AI assistants have no submission process — there is nothing to buy and no lever to pull. A guaranteed first position is generally a term with no competition because nobody searches for it. What can be committed to is the specific work, a recorded baseline of what is currently said, and monthly measurement against a question set that is agreed at the start and not quietly edited later.
How long before any of this shows a result?
Technical fixes can register within weeks. Rankings for anything competitive move over months, and the compounding only begins once several pages are working together. Assistant citations are the least predictable of the three: they can shift after a single model update, in either direction, without anything on the site having changed. This is why the measurement is monthly against a fixed set rather than a single review at the end — a slow trend is readable, and a one-off snapshot is not.
Does schema markup improve rankings?
Not directly, and anyone selling it as a ranking factor is overstating it. What it does is make the relationships on a page explicit — this organisation, this service, this question and this answer — so that a search engine or a language model reads what the page means rather than inferring it. That matters far more for being quoted than for being ranked. The one hard rule is that the markup must say what the visible page says: markup asserting something the page does not display is treated as a quality problem and it is also exactly the inconsistency AEO is trying to remove.
Do we need to write a blog?
You need pages that answer questions buyers actually ask, which is not the same thing. A blog is one container for that and it is frequently the wrong one, because posts date, get filed by month rather than by topic, and compete with the service pages that should own those terms. The usual better answer is a smaller number of permanent pages, each owning one question, kept current — plus a publishing cadence the business can genuinely sustain rather than an ambitious one that stops after two months and leaves a visible date on the last post.