AEO and GEO
Answer engine optimization for B2B
Answer engine optimization is the work of getting your company cited when an AI assistant answers a buying question. DemandBox runs it as a measured programme: a fixed prompt set, a citation baseline, structural and entity work on the pages that should be cited, then repeat measurement against the same prompts.
- Surfaces
- ChatGPT, Perplexity, Gemini, AI Overviews
- Baseline
- Fixed prompt set, measured before and after
- Work
- Structure, entity, schema, source coverage
- Reported as
- Citation share against named competitors
What this covers
An answer engine does not return ten links. It composes an answer and names a handful of sources. Answer engine optimization is the work of being one of those sources for the questions your buyers ask. It overlaps with SEO because both depend on being crawlable and credible, and it diverges from SEO because ranking first is neither necessary nor sufficient for being quoted.
GEO, generative engine optimization, is the same idea aimed at generative results specifically. In practice the work is one programme, and splitting the budget by acronym is a vendor convenience rather than a real distinction.
The failure modes this exists to fix
- A site that ranks well but is never quoted, because its pages restate consensus instead of stating a fact first
- Schema added everywhere with no change in citation, because schema helps parsing and does not persuade a model to prefer a source
- An entity that reads differently across the site, directories and schema, so a model has no consistent description to resolve
- No fixed prompt set, so 'AI visibility' is reported as a feeling rather than a measured share
- Content optimised at the sentence level when the studies show optimisation works at the feature level: structure, source type, evaluative depth, claim specificity
Who this is for
- B2B SaaS companies with commercial and educational pages already live
- Teams who can publish original data, benchmarks or a documented method
- Companies willing to be measured on citation share against named competitors, not a vague sentiment
Who this is not for
- A company with no commercial content for the crawl to find in the first place
- Anyone wanting a guaranteed citation on a fixed date
- Anyone who wants schema added and left there as the whole programme
Verdict: If the site has nothing original to say yet, that gets built first. AEO cannot cite a page that says nothing new.
How the engagement runs
The AEO loop
Prompt set
30 to 60 fixed buyer questions
Baseline
Citation share measured across engines
Structure and entity
Answer-first rewrites, consistent entity
Publish original material
Data, benchmarks, a documented method
Re-measure
Same prompts, same cadence
- 1
Days 1 to 30: build the prompt set and baseline
Write thirty to sixty questions a real buyer would type, covering category definition, comparison and problem framing. Run them across the major answer engines and record who gets cited, how often, and for what. This produces a citation share against named competitors instead of a feeling.
- 2
Days 31 to 60: fix structure and entity
Rewrite the openings of the pages that should be cited so the first forty to sixty words answer the heading outright, since that is the chunk retrieval lifts. Reconcile the entity description across the site, schema, directories and profiles so consensus about what the company is stops reading as noise.
- 3
Days 61 to 90: publish original material and re-measure
Ship the first piece of material that is genuinely yours: a benchmark, a dataset, a documented method. Re-run the original prompt set and report the movement in citation share, not a general impression of visibility.
What you get
Included in an AEO engagement
- ✓A fixed prompt set built around your category and buyers
- ✓A citation baseline across the major answer engines, with competitor citation share
- ✓Structural rewrites of the pages that should be cited and are not
- ✓Entity and schema consistency work across the site and off it
- ✓A plan for the proprietary material that makes you quotable
- ✓Repeat measurement against the same prompts, reported on a fixed cadence
- ✓A monthly review meeting covering movement, not just activity completed
How it is measured
Citation share
citation share = (prompts where you are cited / total prompts in the fixed set) x 100
- prompts where you are cited: count from the fixed prompt set, per engine
- total prompts in the fixed set: held constant between measurement rounds
Example: If 18 of 45 prompts cite you this round against 9 last round, citation share moved from 20% to 40% on the same question set.
| Signal | What it tells you | Source |
|---|---|---|
| 61% of citations to listicle style pages in B2B technology services | Comparison and list formats are heavily favoured by current retrieval | deltavCitations |
| 25,337 citations across 21,075 AI responses, eight industries | Scale of the underlying dataset behind the listicle finding | deltavCitations |
| Under 2% of B2B referral traffic came from AI engines | Traffic from citations is still small; citation share is the leading indicator, not the traffic line | octane11 |
| 91.3% of traffic across 53 B2B SaaS brands still came from organic search | Organic search remains the larger channel; AEO complements it rather than replacing it | piperocketSaas |
What we will not do
- Promise a citation on a fixed date
- Report 'AI visibility' as a sentiment score with no fixed prompt set behind it
- Sell schema markup as the whole programme
- Claim traffic gains from AI engines that current attribution cannot actually show
- Rewrite content at the word level and call it optimisation
Objections, answered
The strongest case against this
AI search sends almost no traffic, so why spend on it now?
Because the citation is doing work whether or not the click happens, and because under 2% of B2B referral traffic currently coming from AI engines is a starting point, not a ceiling. A buyer who reads your name in an AI answer arrives later as a branded search or a direct visit, which is exactly the traffic that looks unattributable in analytics. Waiting until the click volume is obvious means arriving after the consensus about your category has already formed.
The strongest case against this
ChatGPT is the only answer engine that matters.
ChatGPT's share of AI referrals fell from 86% to 72% while Gemini grew 191% and Claude grew 197%, with Gemini now at 13% of referrals. A programme built around one engine is already exposed to that shift. The prompt set and measurement here run across engines for that reason.
The strongest case against this
We will just keep publishing more content and it will follow.
Publishing volume is a token level habit from SEO. The 252,000-trial study behind current AEO understanding found feature level factors, structure, source type, claim specificity, decide citation, not word count. More pages that restate the same consensus do not move a citation share number.
Common questions
- What is answer engine optimization?
- It is the practice of making your content the thing an AI assistant cites when it answers a question in your category. It covers content structure, entity consistency, schema, and being the original source of facts, and it is measured by citation share against a fixed prompt set.
- Is AEO different from SEO?
- They share foundations and diverge on what wins. SEO optimises for a ranked list of links. AEO optimises for being quoted inside a composed answer, where position one is neither required nor enough. In practice one programme should cover both.
- How do you measure whether AEO is working?
- With a fixed prompt set run before and after the work, recording how often you are cited and how that compares with named competitors. Anyone reporting AEO progress without a repeatable prompt set is reporting a feeling.
- Does schema markup make AI cite us?
- No. Schema helps a machine parse a page correctly, which is necessary but not persuasive. Citation follows from clear answers, a consistent entity, and being the source of information rather than a restatement of it.
- How long does AEO take?
- Structural changes can show up in answers within weeks because the crawl cycle is fast. Shifting the consensus about who the credible source in a category is takes longer, and depends on what original material you are able to publish.
- Can you do AEO without doing SEO?
- Not sensibly. An answer engine still has to crawl, index and trust the page. If the technical foundations are broken, the AEO work has nothing to stand on.
- Which AI engines do you measure?
- ChatGPT, Perplexity, Gemini and Google AI Overviews, run against the same fixed prompt set so movement is comparable across engines rather than reported per platform in isolation.
- Do you guarantee citations?
- No. Nobody controls model output directly. What is guaranteed is the process: a baseline, structural and entity work, original material, and repeat measurement. The result is reported honestly whichever direction it moves.
- Does AEO replace paid media or SEO in the budget?
- No. Organic search still carried 91.3% of traffic across 53 B2B SaaS brands in the most recent measurement, well ahead of all AI engines combined. AEO sits alongside SEO and paid media, not instead of them.
Read the thinking behind it
Guides
Want this run for you?
Tell us what you are spending and where the pipeline stalls. If this is not the right first step for you, we will say so.