AEO vs GEO vs SEO, and the fourth one nobody names
Three acronyms for overlapping work, and a fourth surface that none of them measure. A plain comparison with the differences that actually change what you do.
Three acronyms describing largely the same work, sold as three disciplines. Here is the honest version, plus a fourth surface that none of them measure and that matters a lot if you sell developer tools.
The definitions
SEO (search engine optimization) tries to rank a page so a person clicks it. AEO (answer engine optimization) tries to be extracted as the direct answer. GEO (generative engine optimization) tries to shape how AI systems describe your brand across their outputs. Agent discoverability tries to get a coding agent to install your product.
The comparison
| SEO | AEO | GEO | Agent discoverability | |
|---|---|---|---|---|
| Goal | Rank as a link | Be the answer | Be described correctly | Be installed |
| Surface | Google, Bing | AI Overviews, snippets, assistants | ChatGPT, Claude, Perplexity | Claude Code, Codex, Cursor |
| Reader | A person choosing a link | A person reading an answer | A person reading an answer | An agent writing code |
| Result | A click | A citation, often no click | A mention | A lock file entry |
| Metric | Position, clicks, impressions | Answer presence, citations | Mentions, share of voice, sentiment | Install share |
| Main lever | Links, relevance, crawlability | Structure and directness | Content plus third party consensus | What is in the repository |
| Age | 25 years | 3 years | 3 years | New |
Where they are genuinely the same
Most of the work.
Clear structure. Answers stated directly rather than buried. Specific numbers with dates. Tables instead of paragraphs for anything comparable. Crawlable HTML rather than JavaScript-rendered content. Third party sources saying the same thing you say. Content that is current.
Do those, and you have done most of SEO, most of AEO and most of GEO at once. This is why the three-acronym framing oversells the difference. There is one content discipline underneath.
Where they genuinely differ
SEO differs in the currency. SEO buys a click. AEO and GEO often buy a mention with no click at all. That changes how you value the work and how you measure it, but not much about how you write.
AEO differs in the shape. Extraction rewards a specific pattern: question heading, then a 40 to 60 word self-contained answer. SEO tolerated burying the answer. AEO does not.
GEO differs in the horizon. GEO cares about how you are described across many answers over time, which puts more weight on third party sources and consistent facts across the web. One good page can win a snippet. It takes a body of consistent sources to shift how a model describes a company.
Agent discoverability differs in kind. This is not a variation of the other three.
Why the fourth one is a different animal
The first three all assume a person reads something and forms an impression. The fourth has no reader and no impression.
A developer opens a project and types "I need a real database for this, something hosted". The agent lists files, opens the package manifest, reads the lock file, decides what fits, sometimes searches, installs a client library, writes a schema, and shows a diff.
Nothing was read by a person until the diff appeared. There was no answer and no citation.
What decides that outcome is mostly not content:
The repository decides most of it. We ran that database request on JavaScript codebases 111 times and one product won all 111. On TypeScript codebases the same product won 48 out of 132. Same words, different extension, different winner.
The incumbent beats the instruction. A sentence in a repository file saying to use a particular email product was followed 16 times out of 18. Install a competing library into the same repository and the same sentence loses two times out of three.
The agents disagree. Claude Code, Codex and Cursor picked different category leaders in 9 of 18 categories.
Some agents barely read the web. Claude Code ran a web search in 1.6% of decision runs. Codex in 53%. Every content lever is worth roughly thirty times more against one than the other.
None of that is visible to an SEO tool, an AEO audit or a GEO dashboard, because none of them put a repository in the test.
What to do about all four
One order of work covers everything.
1. Fix the documentation. It is the shared input. A quickstart that runs when pasted. The current version stated on the page. One name for one thing across product, package, import and headings. Pages that answer symptoms and not only category names. A Markdown twin of every important page.
2. Fix the technical base. Crawlable HTML. AI crawlers allowed by name in robots.txt. JSON-LD for Article, FAQPage and Organization. A real sitemap. llms.txt at the root.
3. Measure the chat surface. Any prompt monitoring tool. It is cheap and it tells you what a person hears.
4. Measure the repository surface. Real agents, real repositories, hundreds of runs per category, split by agent, language and persona. This is expensive and it is where the installs are.
5. Work the repository. Templates, framework integrations, official examples, platform defaults. Nothing else moves the number as much.
The short answer
SEO, AEO and GEO are one content discipline with three measurement frames. Do the work once and count it three ways.
Then, if you sell a developer tool, go and measure the surface that none of them can see.
Common questions
What is the difference between SEO, AEO and GEO?
SEO tries to rank a page so a person clicks it. AEO tries to be extracted as the direct answer. GEO tries to shape how AI systems describe your brand across their outputs. The tactics overlap heavily and the differences are mostly in what you measure.
Are AEO and GEO the same thing?
In practice, close to it. AEO emphasises extraction, GEO emphasises representation, and both are achieved by clear structure, direct answers, current facts and third party citations. If somebody sells them as two products you are buying one thing twice.
Is SEO dead?
No. Classic search still sends a large share of traffic, and AI systems ground their answers in indexed pages, so the crawlable index still matters. What changed is that ranking first no longer guarantees the click, because the answer may be shown without one.
What is the fourth surface?
A coding agent that reads a repository and installs software. It reads no answer and clicks no citation. It is measured by install share, and none of SEO, AEO or GEO covers it.
Which should a developer tool company invest in first?
The documentation, because it is the input to all four. Then measure the two surfaces separately: prompt monitoring for the chat surface, real agent runs for the repository surface.
Where this comes from
Armature ran 5,292 judged sessions with Claude Code, Codex and Cursor inside 51 realistic codebases, and published every run. The numbers on this page come from that work.