AEO vs SEO vs GEO, in one line each: SEO earns a ranked link in search results. AEO earns a citation inside an AI answer. GEO makes your data usable by the agents that act on those answers. SEO competes on the URL, AEO on the passage, GEO on the knowledge object and the endpoint behind it.
You need all three. Luckily, most of the work overlaps.
In May 2026, Google published its own guide and called all of this "still SEO." Inside Google Search, that's mostly right. Outside it, the math changes. In Goodie's GA4 brand panel, ChatGPT's share of AI referrals fell from 89.1% (May to August 2025) to 62.6% (March to April 2026). Claude's rose from 1.4% to 18.5% over the same stretch. Google's guide doesn't govern either of them.
(Full disclosure: Goodie AI is my company. Weigh the source accordingly, and check the method in the report.)
This playbook covers where the three disciplines differ, what carries over from SEO, what doesn't, and the 90-day plan I'd run.
Why AEO and GEO Matter Now: The 2026 Numbers
- Zero‑click is the norm. In 2024, ~58–60% of both U.S. and EU Google searches resulted in no open‑web click. Users either stopped or refined their query—because answers show up on the results page.
- AI answers are mainstream. AI Mode passed 1 billion monthly active users, per Alphabet's Q2 2026 earnings call (July 2026).
- Behavior shifts when an AI summary appears. In Pew's March 2025 browsing sample, 18% of Google searches produced an AI summary. 88% of those summaries cited three or more sources, and the median ran 67 words. Users clicked a result link on 8% of visits with a summary, versus 15% without.
- Search is turning into an action engine. AI Mode now handles tasks like tickets and reservations, comparing options and handing off to partners.
- Gemini adoption is massive. The Gemini app passed 1 billion monthly users in August 2026.
If you measure your brand with rank and CTR alone, you're undercounting reach. Visibility now includes being cited in answers and being callable by agents.
AEO vs SEO vs GEO: Detailed Definitions
- SEO (Search Engine Optimization): Making pages crawlable, indexable and rankable for keyword queries. Your unit of competition is the URL. You win a position in a list.
- AEO (Answer Engine Optimization): Making your facts and passages the ones AI answers quote. That covers Google AI Overviews and AI Mode, ChatGPT search, Perplexity, Claude, Copilot and Gemini. Your unit is the passage. You win a citation.
- GEO (Generative Engine Optimization): Making your structured knowledge and endpoints usable by generative systems and the agents built on them, so they can use your data and act on it. Your unit is the knowledge object (JSON, CSV, a spec sheet, an image) plus an action endpoint. You win the task.
A note on naming: most of the industry uses AEO and GEO interchangeably. Microsoft's January 2026 guide splits them differently again, reading AEO as "agentic engine optimization." I separate them because the work is different. Writing a quotable passage and exposing a booking API are two jobs, owned by two different people.
Google's guidance still starts with helpful, reliable, people-first content. That's the floor for all three.
Mental models: the three pipelines
- SEO: Crawl → Index → Rank. Links, topical depth and page experience decide your position in the list.
- AEO: Retrieved → Cited → Trusted. The engine fans the question out into related sub-queries, pulls passages from its index or the live web, drafts a short answer and cites a handful of sources. Google confirms its AI features use exactly this query fan-out against the Search index. If your page doesn't hold a crisp, verifiable statement, there's nothing to pull.
- GEO: Goal → Reason → Call → Complete. The agent splits a goal into sub-tasks, compares options and calls partner systems to buy or book. If you expose nothing machine-readable, you aren't eligible.
What Carries Over From SEO (and What Doesn't)
Here's the honest version: inside Google, most AEO is SEO done well. The gaps open up the further you get from Google's index.
| SEO practice | Carries over to AEO? | What changes |
|---|---|---|
| Crawlable, indexable pages | Fully | Many AI fetchers don't execute JavaScript. Server-render anything you want quoted. |
| Topical depth | Fully | Judged passage by passage, not page by page |
| Original expertise | Fully, and it matters more | Google's guide names non-commodity content as the factor that counts. Generic listicles lose on both surfaces |
| Backlinks | Partly | Third-party mentions shape answers with or without a link. In Goodie's social citations study, 82.5% of YouTube citations came from Google surfaces and 99.7% of X citations from Grok |
| Keyword targeting | Partly | Prompts are longer and conversational. Fan-out means the sub-questions matter as much as the head term |
| Rank tracking | No | Position 1 means little if the answer omits you. Track citation share by engine |
| CTR and sessions | Partly | Answers are zero-click by design. Add GSC's generative AI report and AI referral sessions in GA4 |
| Schema | Partly | Google says there's no special AI schema. Still worth it for rich results and clean entity data |
SEO is the entry ticket to AEO. It doesn't win the citation on its own.
And no, AEO doesn't cannibalize SEO. The same answer-first page that wins an AI Overview citation usually earns the featured-snippet-style placement too. The real trade-off is budget and headcount, and that's a reporting problem before it's a strategy problem.
AEO vs SEO vs GEO: Side-by-Side Comparison (2026)
| Dimension | SEO | AEO | GEO |
|---|---|---|---|
| Primary surface | Google and Bing results | AI Overviews, AI Mode, ChatGPT search, Perplexity, Claude, Copilot | Agents and agentic modes (shopping assistants, booking flows, custom agents) |
| Unit of competition | URL | Fact / passage | Knowledge object + endpoint |
| Trigger style | 2–5-word keywords | Natural-language questions | Conversations + files/images |
| Result shape | Ranked list | Short cited answer + links | Persistent session + actions |
| Selection signals | Links, topical depth, CWV | Helpfulness, verifiability, freshness, third-party corroboration | Source authority, structured data, API availability |
| Core KPI | Impressions, CTR, sessions | Citation share by engine, AI referral sessions, sentiment | Agentic conversions, task completion |
| Ignore it and… | traffic decays | answers omit you | agents buy/book without you |
| Where you measure it | GSC, GA4 | GSC generative AI report, citation trackers like Goodie, GA4 AI referrals | Partner and agent logs, server logs |
(Yes, SEO still matters. It's just no longer the whole funnel.)
The Playbook: How to Run AEO and GEO Alongside SEO
1. Write in quotable units (AEO)
Ship content in quotable quanta:
- Answer-first openings. A direct answer under 60 words, right under the H1. Google says you don't need to chunk content for AI, and it's right. You do it because every engine, and every human skimmer, lifts the first clean answer it finds.
- Numbered steps and checklists — easy to summarize and cite.
- Evidence lines: “In March 2025, X did Y [source]” — with an external link to a reputable publisher.
- Tight tables (comparisons, specs) with clear headers.
- Standalone FAQs per intent, not massive omnibus pages.
Pew's March 2025 data shows why: the median Google AI summary ran 67 words and 88% cited three or more sources. Give the engine a verifiable claim it can lift.
Micro‑pattern to copy:
Verdict: "[Option A] is the better pick for [use case] because [one verifiable reason]. [Option B] wins if [other criterion] matters more."
Why: three bullets, each with a dated source.
How: three steps, each linking to a deeper guide.
2. Technical accelerators (AEO → GEO)
- Schema where it earns its keep: Organization, Person, Article and Product give engines unambiguous entity data. Add FAQPage where real FAQs live on the page. Don't expect markup alone to earn a citation. Google says there's no special AI schema.
- llms.txt is optional: Google Search ignores it. Across the files we tracked, 97% weren't fetched by anything in a month. It helps coding agents when something points them at it. Publish one if it's free to make: your CMS may generate it, or our free llms.txt generator at Goodie drafts one from your URL in seconds. Don't put a sprint behind it. (Full breakdown in my llms.txt guide.)
- Server-render the answer: Many AI fetchers don't run JavaScript, so content injected client-side may never be seen. Check that your robots.txt and CDN bot settings let AI crawlers in.
- One canonical page per fact type: pricing, specs and policies each live on one page, updated in place with a visible date, so answers don't cite a stale copy. Don't spin up a page per sub-query.
- Versioned PDFs with front‑matter (title, date, author) for credible citations.
3. Turn your site into a knowledge surface (GEO)
- Expose task APIs: booking/reservations, calculators, availability, pricing. That’s how AI Mode “does the work for you.”
- Publish machine‑readable objects: JSON/CSV for specs, fees, sizes; concise alt text and captions so multimodal models parse images correctly.
- Vectorize your docs for your own assistant: it keeps the answers on your site consistent. It doesn't change what third-party models retrieve. Only your public, crawlable facts do that.
Copy‑paste recipes
A. Minimal llms.txt (proposed standard)
# YourBrand Docs & Facts
> YourBrand helps [audience] do [job]. Key pages, pricing and facts are listed below.
## Start Here
- https://example.com/what-we-do: 120-word summary of product & who it's for
- https://example.com/pricing: Current public pricing (updated monthly)
- https://example.com/security: SOC2, data residency, compliance docs (PDFs)
## Fact Sheets
- https://example.com/datasheet.csv: Product specs as CSV (field dictionary at top)
- https://example.com/faq: Top 25 FAQs with one-sentence answers + sourcesIf you publish one, keep it this simple, and expect coding agents to use it, not answer engines. The format comes from Jeremy Howard's proposal at llmstxt.org.
B. FAQPage schema (for pages with real FAQs)

How to Measure SEO vs AEO vs GEO
Rank and CTR still tell you how SEO is doing. They tell you almost nothing about AEO. Two numbers I'd put on every dashboard:
- Answer Share of Voice (A-SoV) = answers citing you ÷ total answers observed for your topics
- Agentic Conversion Rate (A-CR) = agent-initiated completions ÷ agent-initiated sessions
Start with the free data. Google's Search Console now has a generative AI performance report, and Google says it's the only source of its internal AI metrics. Third-party tools can't see them. In GA4, segment referrals from ChatGPT, Claude, Perplexity, Gemini and Copilot.
For everything outside Google, you need a citation tracker. We built Goodie for this, so that's what we use. Compare the category before you buy.
Why bother? Across 117,432 inbound leads at 13 B2B companies (October 2024 to April 2025), Goodie found AI search leads closed at a 56.3% higher rate than Google and Bing leads. That's the case for measuring the surface, even when the session count looks small.
I go deeper on the framework in how to measure AI search visibility.
Teams & workflow: who owns what
| Role | Classic SEO era | 2026 job |
|---|---|---|
| SEO Manager | Rankings & CWV | AEO lead: citation share, freshness SLAs, schema QA |
| Content Lead | Long-form briefs | Prompt-native editor: answer-first formatting, fact boxes, external sources |
| Web Engineer | Templates, sitemaps | Server-rendered answers, AI crawler access, task endpoints |
| Analytics | GSC + GA4 | Visibility scientist: A-SoV, hallucination rate, agentic conversions |
| PR/Comms | Media mentions | Narrative steward: ensure consistent facts across press, site, and AI answers |
The bigger org question is who owns AI visibility when SEO, social and PR all shape the answer. I wrote about that in Rethinking the marketing organization for AI search.
What Google Says, and Where It Stops
Google's May 2026 guide is worth reading in full. Its position: AI Overviews and AI Mode run on the same core ranking and quality systems as Search, so optimizing for them is still SEO. It calls llms.txt, chunking and special schema unnecessary.
Inside Google, I mostly agree. But the guide also confirms the retrieval layer changed: query fan-out and retrieval-augmented generation against the index. And it speaks for Google only. ChatGPT, Claude, Perplexity and Copilot retrieve, weight and cite on their own terms.
- AI Overviews: short, cited answers. Be the source that's easiest to verify.
- AI Mode: passed 1 billion monthly users. Deeper research plus tasks. Prepare facts (AEO) and action endpoints (GEO).
- Everything else: measure it separately. A Google-only view will undercount you. My full take: Did Google just kill AEO/GEO?
Common failure modes (and fixes)
- Thin claims with no sources → Add footnotes and respected external citations. Snapshot writers reward verifiability.
- Bloated pages → Front-load the answer (under 60 words), then expand.
- All text, no structure → Add JSON‑LD + concise tables.
- No API → Expose calculator/booking/availability endpoints; AI Mode can only act if you’re callable.
- No machine-readable facts → Publish one canonical, dated page per fact type (pricing, specs, policies).
AEO vs GEO Tactics by Search Intent
| Intent | AEO tactic | GEO tactic |
|---|---|---|
| What is… | 40-word definition box + 3 sources | Definition + JSON dictionary endpoint |
| How to… | Numbered steps + one image per step | Steps as JSON + short clip; expose checklist API |
| Best X for Y… | Clear picks with criteria table + price range | Comparison dataset (CSV/JSON); stock/pricing API |
| [Brand] vs [Brand] | Pros/cons table, test methodology, citations | Side-by-side spec endpoint agents can query |
| Pricing & plans | Transparent table + billing rules | Pricing API with version tags; return JSON & CSV |
| Local service | Service area & NAP in LocalBusiness | Booking/availability + calendar integration |
30‑60‑90 rollout (no–nonsense)
A. Days 0–30 (Audit & Patches)
- Add an answer-first verdict and an on-page FAQ to your top 20 money pages. Mark up the FAQ.
- Check AI crawler access: robots.txt, CDN bot rules, server-rendered content.
- Create a Facts page with timestamped claims + external sources.
- Turn on the GSC generative AI report and an AI-referrals segment in GA4.
B. Days 31–60 (Vectorize & Structure)
- Build a mini knowledge base: embed docs, image captions, tables.
- Stand up a GraphQL/REST endpoint for specs/pricing/availability.
- Instrument A-SoV tracking across AI Overviews, ChatGPT, Perplexity, Claude and Gemini.
C. Days 61–90 (Agentic Enablement)
- Expose at least one task API (demo booking, calculator, reservation).
- Add hallucination monitors (spot wrong facts in snapshots) and fix upstream pages.
- Publish a model‑ready press kit (mission, leadership bios, datasets) linked from your About page and sitemap.
A note on: memory, alignment, faithfulness
An LLM’s “memory” of your brand is a by‑product of what it can retrieve now + what it learned earlier. If your public facts are inconsistent, slow, or buried, answers will drift. You fight that with fresh, citable facts (AEO), consistent entity data (schema), and authoritative endpoints (GEO). That’s how you improve alignment (answers reflect your intent) and faithfulness (answers match reality).
Where Goodie AI Fits
We built Goodie because the AEO column of that table had no good instrument. It does four jobs:
- Monitor: citations, share of voice and sentiment across every major engine, including ChatGPT, Claude, Gemini, Perplexity, Copilot and Google's AI surfaces. It also shows the prompts where competitors get cited and you don't.
- Optimize: a prioritized list of what to fix or publish to close those gaps.
- Agent analytics: how AI crawlers actually experience your site.
- Attribute: which AI platforms send traffic, and what that visibility is worth in revenue.
Then compare it against the rest of the category before you buy. My take on the options is in the AEO tools roundup.
If you need one page to hand your exec team, start with my expert guide to AEO.
The takeaway
- SEO gets you found.
- AEO gets you quoted.
- GEO gets you used.
Search went from lists to answers, and it's moving to actions. Google can call the first two "SEO" and be mostly right about its own surfaces. The third doesn't care what you call it. If your data isn't callable, the agent books with someone else.
Curious what you're seeing in your own data.
