Search engine optimisation decides which page a person clicks. Answer engine optimisation decides which source a model quotes when it answers instead of listing. They share an indexing layer and almost nothing else.
Kunzum measured this across 1,500 AI responses in September 2026 — 480 in the Crypto AI Visibility Index and 1,020 in the Crypto Card Report — logging every citation each engine produced. The short version of what that showed: the mechanism SEO optimises and the mechanism AEO optimises come apart most of the time, and a page can win one while losing the other.
In short
- SEO ranks a page. AEO gets a passage quoted. A buyer reading an AI answer may never see a ranked list at all.
- Kunzum’s September 2026 index logged 12,518 citations across 1,396 domains from 1,500 responses. The most-cited domain in both editions was reddit.com, at 990 citations — ahead of every vendor, publisher, bank and regulator.
- Retrieval is not guaranteed. ChatGPT ran a web search on 54% of its answers in edition one and Claude on 65%. The rest came from training data, where nothing published could have reached them.
- Google states there is no AEO-specific markup. Anyone selling schema as the mechanism for AI Overviews is selling something Google has published a denial of.
- Neither replaces the other. If your buyers still search, SEO still pays; AEO is the half that has no impression log.
What actually differs
The distinction is not a matter of emphasis. Kunzum treats them as different jobs because the thing being optimised is different: SEO competes for a position in a list, AEO competes to be the sentence a model reuses. Everything downstream follows from that.
| SEO | AEO | |
|---|---|---|
| Unit of success | A URL ranked on a results page | A passage quoted inside an answer |
| What the buyer sees | Ten links and a choice | One answer, sometimes with three citations under it |
| What you change | Titles, internal links, depth, speed | What a paragraph claims, and whether it survives being read alone |
| Off-site work | Links pointing at you | Being the source other pages state facts from |
| How you know it worked | Rank position, impressions, clicks | Ask the engine the question again and see whether you are named |
| What analytics records | Impressions and sessions | Nothing, unless the reader clicks a citation |
| Characteristic failure | Page two | Ranking first on Google and being absent from the answer above it |
The row that matters commercially is the last one. A crypto protocol ranking first for its category term, with an AI Overview above it naming three competitors, has an SEO problem no SEO report will show, because by every SEO measure that page is winning.
Where the two genuinely overlap
AEO is not a separate internet, and it is worth being precise about the overlap rather than overselling the difference. As of September 2026, Google’s Search Central documentation on AI features states that to appear in AI Overviews or AI Mode “a page must be indexed and eligible to be shown in Google Search with a snippet”, and adds: “You don’t need to create new machine readable files, AI text files, or markup to appear in these features. There’s also no special schema.org structured data that you need to add.”
Two things follow. Technical SEO still matters, because an unindexed page is invisible to both. And the large category of AEO services built on structured data is selling a mechanism the largest engine has publicly denied. Structured data helps a machine parse a page it already has; it does not cause retrieval.
The engines that are not Google describe their own pipelines the same way. OpenAI’s crawler documentation separates OAI-SearchBot, used “to surface websites in search results in ChatGPT’s search features”, from GPTBot, which gathers training data, and Perplexity’s crawler documentation states PerplexityBot “is not used to crawl content for AI foundation models”. Four engines, four doors, and a robots.txt tuned for Googlebot governs none of them.
What the measurement showed
Two numbers separate AEO from SEO more clearly than any argument. The first is how often an engine goes to the web at all. In edition one of the Kunzum index, across 480 responses:
| Engine | Answers that ran a web search | Citations per answer |
|---|---|---|
| ChatGPT | 65 of 120 (54%) | 3.6 |
| Claude | 78 of 120 (65%) | 2.7 |
| Perplexity | 120 of 120 (100%) | 19.3 |
| Google AI Overviews | Always retrieves | 7.7 |
An answer produced without retrieval cannot have been changed by anything anyone published. That is the ceiling on this entire field, and it is why the honest claim is that AEO moves the retrieved half rather than all of it. It also explains why engines have to be treated separately: Perplexity cited 19.3 sources per answer and Claude 2.7. Work that satisfies one does not automatically reach the other.
The second number is where the citations went. Across both editions, 12,518 citations landed on 1,396 domains, and reddit.com was the single most-cited domain in each — 293 citations in edition one and 697 in edition two. In edition one, vendor pages took 3,409 citations and editorial press took 26. A backlink campaign aimed at trade press is optimising a channel that supplied under one percent of what these engines quoted.
The full citation breakdown, by engine and by source type, is published with the raw data behind it.
When SEO is the better place to spend
Kunzum sells AEO, so this section is the one worth weighting. There are cases where answer-engine work is the wrong priority and saying so costs nothing.
If your buyers still convert through search — if Search Console shows real commercial queries bringing real signups — that channel is measurable, attributable and already working, and AEO is an addition to it rather than a replacement. If your site is not indexed cleanly, AEO is premature, because the same indexing layer gates both. If you are pre-product, or the category genuinely has no buyer questions yet, neither is the problem.
AEO earns priority when the questions your buyers ask are being answered by a model that names other people, and you cannot see it happening. That is a specific, checkable situation rather than a general claim about the future, and the method starts by checking it.
Two narrower reads sit alongside this one: what Google has actually published about AI Overview eligibility, which rules out several popular line items, and the three different services sold as AI SEO, which is worth reading before comparing two proposals.
Sources
- Google Search Central, “AI features and your website”. Google. Checked 2026-09-12.
- OpenAI’s crawler documentation. OpenAI. Checked 2026-09-12.
- Perplexity’s crawler documentation. Perplexity. Checked 2026-09-12.
- Aggarwal and colleagues, “GEO: Generative Engine Optimization” (2023). arXiv:2311.09735. Checked 2026-09-12.
- Liu, Zhang and Liang, “Evaluating Verifiability in Generative Search Engines” (2023). arXiv:2304.09848. Checked 2026-09-12.
Published 2026-09-12. Written by Narender Charan, who runs Kunzum.