AEO vs SEO: What Changes, What Stays and What to Fix First
AEO sits on top of SEO. Here’s which work is the same, which has shifted, and which is new, using the platforms’ own documents, plus a ranked list of what to fix first.
AEO vs SEO is a sequencing question, not a replacement one. SEO makes pages findable and trusted, and AEO shapes those same pages so an answer can be lifted from them. Fix the foundation first.

One person on your team forwarded you a post saying search optimization has been replaced. Someone else sent one saying nothing has changed.
Both are selling something.
So am I. We sell SEO and AEO work, so weigh what follows accordingly, and check every platform claim against the pages linked here, which I read this week.
My view is that you have one system, not two. The pages that earn an AI answer are the pages that earn a ranking, with a few changes at the edges, and if you get the order of the work wrong you end up paying to polish pages that no crawler can reach.
The Short Version
AEO sits on top of SEO. Google says optimizing for its AI features is still SEO, while the other assistants’ documents cover crawler rules and not ranking rules. So rank the jobs by the damage if each one is broken and by how cheap it is to check: access, crawler settings, the pages that sell, how those pages open, then schema and measurement. Our answer engine optimization services page shows how we run the AEO layer. The neighboring pairs have their own posts, AEO versus GEO and GEO versus SEO.
AEO vs SEO: What Each One Is, and Where They Overlap
SEO is the work of making your pages crawlable, indexed, relevant, and trusted, so they can rank in search results. AEO, answer engine optimization, is shaping those same pages so a search surface or an assistant can lift a short, correct answer from them. One earns the place. The other earns the quote.
The overlap is bigger than the labels suggest. Google says so in its guide to generative AI features: “From Google Search’s perspective, optimizing for generative AI search is optimizing for the search experience, and thus still SEO.” That’s Google talking about Google.
Treat it as strong evidence for one engine and weak evidence for the rest. OpenAI, Anthropic, and Perplexity publish rules for their crawlers. In the pages I read for this piece, none of them describes how a page gets ranked. So the table sorts each job by what’s documented, and says whose document it is.
Same, Shifted or New?
| The work | Verdict | What the platforms say |
|---|---|---|
| Crawlable, indexed pages that return a 200 | Same | Google: a page must be indexed and eligible to show with a snippet to appear in its AI features. OpenAI: a site opted out of OAI-SearchBot isn’t shown in ChatGPT search answers. |
| Unique content with a point of view | Same | Google: non-commodity content will likely matter to its AI features in the long run more than any other suggestion in its guide. |
| Structured data that matches the page | Same | Google: not required for its AI features, and no special markup exists. Keep it accurate and matching the visible text. |
| How a page opens and is organized | Shifted | Bing: clear headings, tables, and FAQ sections help AI systems reference content. Google: write for readers, with no need to chunk. |
| Crawler access, bot by bot | New | OpenAI, Anthropic, and Perplexity each run separate bots for search, training, and user requests. |
| Freshness signals | Shifted | Bing: the sitemap lastmod field and IndexNow shape how quickly updates reach search results and AI answers. |
| What you measure | New | Google counts AI features inside Search Console’s Web search type and points to a Generative AI performance report. Bing Webmaster Tools reports citations. |
| llms.txt and other AI-only files | Optional | Google: its Search ignores them, so they neither help nor harm there. |
Three of the eight rows are the same job as before. Two have shifted. Two are new. One is optional. That’s your budget conversation in a single table.
AEO vs SEO: What Stays the Same
Start with the plumbing. Google says its generative AI features rely on retrieval-augmented generation, which pulls relevant pages from its Search index to ground an answer, and on query fan-out, where the model runs a set of related searches at once to answer one question. Its own example is a question about fixing a lawn full of weeds, which fans out into searches on herbicides, chemical-free removal, and prevention.
Translate that to your world. A buyer asks about co-managed IT, and the model also checks onboarding, pricing models, and security tooling. A page that covers the neighboring questions has more ways to get picked up. That’s content depth. It’s SEO work, and it was before anyone said AEO.
Then the eligibility rule. To show in Google’s AI features, a page must be indexed and eligible to appear in Search with a snippet. A noindex tag left over from a redesign. A blocked folder. A page that returns an error. No AEO tactic fixes any of them. Google’s page on AI features and your website adds that there are no additional requirements to appear in AI Overviews or AI Mode. It’s dull, and it’s the cheapest thing in this post to check.
Then the content itself. Google asks for unique, “non-commodity” work, and says it will likely influence a site’s presence in generative AI search in the long run more than any other suggestion in the guide. Its example of commodity content is “7 Tips for First-Time Homebuyers,” because it’s built on common knowledge that anyone could have written. The MSP version is “10 Cybersecurity Tips for Small Business.” Every competitor has that page. So does every model.
What’s non-commodity? Your actual onboarding process. Your terms. The jobs you turn down, and why. A point of view a buyer can disagree with.
Thin pages don’t become citable because somebody added a schema block.
What Changes: The Click, the Crawlers, and the Opening of the Page
Three things move. None of them is a new craft. All of them change what you check.
The click stops being the scoreboard
Pew Research Center tracked the browsing of 900 U.S. adults in March 2025. When a Google results page showed an AI summary, people clicked a regular result link in 8% of visits. Without one, 15%. That’s a small panel: one month, U.S. only, Google only. Take the direction, not the decimals.
Google tells it differently. Its May 2025 post says clicks from AI Overviews are higher quality, with visitors more likely to spend time on the site, and it tells owners to look at conversion indicators and not at clicks alone. That’s Google describing its own product, so read it as a claim, not a finding.
Both readings land in the same place. Judge the work by what reaches your pipeline, which is the argument in SEO KPIs that tie to pipeline. A falling click line and a rising call log can both be true in the same quarter.
More bots, and each has its own switch
The assistants split their work across several bots. OpenAI, Anthropic, and Perplexity each run separate bots for search, for training, and for fetching a page when a person asks. The table below comes from each company’s own page (OpenAI’s crawler overview, Anthropic’s Help Center article, and Perplexity’s crawler guide), and those pages change, so check the source before you edit a file.
| Bot | What it’s for | What its owner says about blocking it |
|---|---|---|
| Googlebot and the Search index | Feeds Google’s AI Overviews and AI Mode | A page must be indexed and snippet-eligible, and nosnippet and max-snippet limit how it’s featured in AI formats. |
| OAI-SearchBot (OpenAI) | Surfaces sites in ChatGPT search | Opted-out sites aren’t shown in ChatGPT search answers, though they can still appear as navigational links. |
| GPTBot (OpenAI) | Collects content that may train OpenAI’s models | Disallowing it signals that the content shouldn’t be used for training. |
| ChatGPT-User (OpenAI) | Visits a page when a ChatGPT user asks | Robots.txt rules may not apply, and it isn’t used to decide what appears in search. |
| Claude-SearchBot (Anthropic) | Improves the quality of Claude’s search results | Disabling it may reduce your visibility and accuracy in search results. |
| Claude-User (Anthropic) | Visits a page when a Claude user asks | Disabling it stops retrieval in response to a user’s question and may reduce visibility. |
| ClaudeBot (Anthropic) | Collects content that may train Anthropic’s models | Restricting it signals that future material should be left out of training data. |
| PerplexityBot (Perplexity) | Surfaces and links sites in Perplexity search results | Not used to crawl for foundation models, and Perplexity recommends allowing it to appear in search results. |
| Perplexity-User (Perplexity) | Fetches a page when a Perplexity user asks | Because a person requested the fetch, it generally ignores robots.txt rules. |
Read the pattern. Every company separates search from training. So a blanket “block AI” rule, the kind someone adds after a security scare, can shut out the bots that decide whether you appear in an answer. Whether to allow training is a legal and ownership call. Whether to allow search is a revenue call. They’re different decisions, with different owners.
Our post on whether to block AI crawlers walks through both calls. Note the two fetchers that act when a person asks for a page. OpenAI says robots.txt rules may not apply to ChatGPT-User, and Perplexity says Perplexity-User generally ignores them, while Anthropic says all three of its bots honor robots.txt. A robots.txt file is a request. Not every fetcher treats it as one.
The opening of the page
AEO in its narrow sense lives here. Put the answer first, where the reader and the extractor both look.
Microsoft’s announcement of AI Performance in Bing Webmaster Tools says clear headings, tables, and FAQ sections help AI systems reference content accurately, and that examples, data, and cited sources build trust when content is reused. Google says something that sounds like the opposite: there’s no requirement to break content into tiny pieces for AI, and no need to write in a special way for generative AI search. Both can be true. Write for the reader, put the answer where the reader looks first, and don’t slice the page into fragments for a machine.

One habit to drop. Building a separate page for every variation of a question, mainly to move rankings or AI answers, is what Google’s guide says violates its scaled content abuse spam policy. It adds that the tactic doesn’t work over time, because a high quantity of pages doesn’t make a site more relevant.
Freshness gets one more line. Bing says the lastmod field in your sitemap is a key signal for what it recrawls, and it recommends IndexNow so participating engines learn when content is added, updated, or removed. A sitemap that stamps the same date on every page isn’t telling anyone anything.
AEO vs SEO: What to Fix First, Ranked by Risk
Here’s my rule. Rank each job by the damage if it’s broken, then by how cheap it is to check. Price the risk first. It’s a working view, not a Google rule and not a benchmark, and it’s the order I’d give an MSP owner whose proposal pile holds an SEO retainer and an AEO retainer.
- Access first. Can Googlebot reach every page that sells, and is each one indexed and showing a snippet? Search Console tells you. A stray noindex or a blocked folder is the costliest mistake on this list and the cheapest to find.
- Then read your robots.txt line by line against the table above. Decide search and training separately, and write down who decided.
- Service, pricing, and comparison pages come next. They need what a model can’t get from every other MSP site: your process, your terms, your limits. Thin pages get fixed here, not at step four.
- The opening is where AEO in its narrow sense starts. Rewrite the first 40 words of those pages so they answer the buyer’s question outright. It’s quick, and it’s worthless if the first three steps are broken.
- Schema and sitemap dates. Keep structured data accurate and matching what’s visible on the page. Fix the dates. Switch on IndexNow if your platform supports it.
- Last, measure. Search Console for Google, Bing Webmaster Tools for Copilot, a log of buyer questions you ask by hand in ChatGPT, Claude, Perplexity, and Gemini, and one field on every first call: “How did you find us?” The setup is in our post on measuring AI search visibility.
Three platforms get their own pages from us: ChatGPT SEO services, Bing SEO services, and Claude SEO. For a page-by-page pass, use our AEO checklist.
Look at what’s missing. No new content for every query variation. No separate “AI site.” Nothing with AI in the product name. A broken foundation costs the same before and after you add an answer layer, except that afterward you’ve paid for the layer too, and you’ve spent the months you could have used on the thing that was actually broken.
Off-site work comes last. It’s the slowest, the hardest to see, and it depends on other people. When your pages are open, indexed, and answer-first and the assistants still skip you, you’ve reached the territory covered in the split between AEO and GEO and in our look at brand mentions and AI search. How we run all of it together is on our page for SEO, AEO and GEO for tech companies.
What Not to Pay For on Either Side
- A file sold as the unlock. Google says its Search doesn’t use llms.txt or other special AI files, and that having one neither helps nor harms there. Keep one if another service you care about reads it. Don’t buy it as a strategy.
- A tool with inside knowledge. Google’s guide warns that no third-party tool has access to its internal ranking or AI systems, and says to weigh any tool’s advice against its official guidance. Use the tools if they speed up your work. Don’t mistake their score for Google’s.
- Mentions bought in bulk. Google says seeking inauthentic mentions across the web isn’t as helpful as it might seem.
- A date. If someone promises when an assistant will name you, ask how they’d know. Nobody outside the model does.
Sequence the work, then spend the money. Not the other way around. If a proposal on your desk splits SEO and AEO into two retainers, bring it, and we’ll read it with you against the table above. Book a growth strategy call and send it ahead.
What Owners Ask Before Splitting the Budget
Is AEO replacing SEO?
Not on Google’s own account. Its guide says optimizing for generative AI search is still SEO, because the features are rooted in its core ranking and quality systems.
The other assistants’ documents don’t describe ranking rules, so nobody outside them can say what replaces what. Plan for both, and start with the shared foundation.
Do we need two budgets or one?
One plan, with no more than two lines in it. Access, content, and structure are shared work, so two retainers can end up paying for the same audit twice.
Give AEO its own line only for page openings and measurement, and only once the foundation is clean.
Does ranking well on Google get us into ChatGPT answers?
Treat it as unknown and test it. Google says its AI features draw on its core ranking systems, but the OpenAI page I read documents its crawlers and not how it picks sources.
Ask ChatGPT the questions your buyers ask, and write down which companies come back.
Should we add an llms.txt file?
Optional, and not for Google. Google says its Search ignores llms.txt, so the file neither helps nor harms there.
If another service you care about reads it, maintain it. Otherwise spend that hour on page access.
How do we know the AEO work is paying off?
Three places. Search Console for Google’s AI features, the AI Performance report in Bing Webmaster Tools for Copilot citations, and the answer to “How did you find us?” on first calls.
A citation count proves a page was used. Only the call log shows whether it was worth anything.
Can a small team run both?
A small team can, if one person owns it. The first four steps need access to robots.txt, Search Console, and a few hours.
Where teams stall is ownership, not skill. Name the owner, then put a date on the calendar.
Find out which work your site needs first
Talk to us about whether your pages are open, indexed, and answer-first, so you fund the right work in the right order.
Book a Growth Strategy Call