How to optimize for AEO in 2026: most of what's being sold is repackaged SEO. Here's the founder playbook for getting cited in ChatGPT.

I run an AEO agency, which means I spend most of my week figuring out why some B2B SaaS brands get cited inside ChatGPT, Claude, and Perplexity when buyers ask "what's the best X for Y," and others don't. Most of what's being sold as answer engine optimization right now is repackaged SEO with new vocabulary. The playbook that actually moves citations looks almost nothing like the playbook that moved Google rankings, and the founders winning in AI search figured this out about eighteen months before their competitors.

Here's how to optimize for AEO the way I use it with clients. Six steps. No SEO templates. No "write more content" advice.

But first, the mental model that has to come first, because if you skip this you'll do the work and get nothing.

The shift you have to internalize

Google ranks pages. LLMs cite passages.

That sentence is the entire difference between SEO and AEO, and most agencies are still optimizing for the first one.

When somebody asks Google "what's the best HR software for mid-market," Google returns ten blue links and the searcher clicks one. The agency's job is to win the click. When somebody asks ChatGPT the same question, the model returns a three-name shortlist with a confident summary. The user never clicks anything. The agency's job is to be one of the three names.

The mechanics of how those three names get picked are not the same as the mechanics of how Google ranks. Citation is a retrieval-plus-generation problem. The model has to pull a passage from somewhere, attribute it to a brand, and decide that brand is one of the top answers to the question. Each of those steps has its own optimization vector.

The playbook below is built around those mechanics, not around the SEO assumptions that don't translate.

Step 1: map the question taxonomy your buyers actually ask

Stop with the keyword research. Keywords are what people typed into Google when they had to compress their question into three words. Questions are what people ask ChatGPT when they don't have to.

Sit down with your sales team and your customer success team. Write out the actual questions your buyers ask during evaluation. Not their search queries. Their actual questions.

For an HR software company, this looks like:

• "What's the best HR software for a 50-person remote SaaS company?"

• "Which HRIS handles US and EU compliance without a separate EOR?"

• "What's the cheapest way to run payroll across 5 countries?"

• "Is Rippling or Deel better for a company doing 70% contractors?"

These are full sentences. They have context. They name competitors. Your buyer is having an actual conversation with the model, and the model is treating each question as its own retrieval problem.

Build a list of 30 to 50 of these for your category. They are the questions you need to be cited for, and they're nothing like the keyword list your SEO agency produced.

Step 2: measure your baseline citation share

You can't optimize what you're not tracking. Before you change anything, query each of the major answer surfaces you're tracking (ChatGPT, Claude, Perplexity, Gemini, Google AI Overviews) with each of your 30 to 50 questions, and record whether your brand is cited.

Three outcomes per question per answer surface:

• Cited by name as a recommended option

• Mentioned in passing without recommendation

• Not present

The only outcome that matters is the first one. Mentions don't move pipeline. Recommendations do.

In the baselines I've run for clients, most B2B SaaS brands start well below 50% citation share across the high-intent questions in their category the first time we measure. That's the gap to close. Without the baseline, you have no way to know whether anything you're doing later is working.

Step 3: rewrite your owned content as answer-shaped, not blog-shaped

LLMs don't read your page top-to-bottom. Retrieval systems scan for passages that answer the question they're processing. Treat a useful extractable passage as roughly 50 to 200 words. It needs to contain the answer in the first sentence, followed by the supporting context.

A blog post structured as "intro, headers, paragraphs, conclusion" is hard to extract from. A post structured as "question, direct answer, context, evidence" is easy.

The rewrite for AEO looks like this:

• Lead each section with the question it answers, as a real-language question

• Open the section with the direct answer in one or two sentences

• Follow with the context, evidence, and qualifications

• End with the next question this section sets up

This is how to write content the engines can lift verbatim. Most SEO content fails this test because it was written for a reader scrolling, not a model extracting.

If your existing content already ranks on Google, you don't have to throw it out. You restructure it. In my experience, a passage-shape rewrite can create a meaningful lift in citation share without any new content being created.

Step 4: build the third-party signal layer

This is the step nobody else is doing seriously, and it's where most of the citation lift comes from.

The engines decide whether to recommend your brand by looking at how third-party sources describe you. If five independent "Best X for Y" articles list your brand among the top three, the model learns that your brand is one of the top three. If those same articles never mention you, the model learns you're not in the consideration set.

Building this layer is operational work. You identify the third-party surfaces the engines pull from for your category. Industry roundups, UGC sites like Reddit, Indie Hackers, Hacker News, and LinkedIn, comparison sites, review platforms like G2 and Capterra, specific high-authority blogs in your vertical. You make sure your brand is present on each of them, accurately described, and ideally ranked well.

That does not mean fake reviews, planted UGC, or doorway pages. That's spam with a new costume.

This is not link building. Link building was a domain-authority play. This is mention building. The link doesn't matter. The mention does.

Most of the agencies selling AEO right now are skipping this step because it's harder to staff than blog content. It's also the step that separates the founders who win at AI search from the ones who lose.

Step 5: get the structured data right

Structured data gives crawlers and search systems cleaner facts to associate with a brand. Good schema markup can say "this product is made by this company, costs this much, integrates with these tools, serves this use case."

Minimum schema set for a B2B SaaS company doing AEO, used only where the page actually supports it:

• Organization schema on every page (consistent name, logo, description, social profiles)

• SoftwareApplication schema on product pages, or Product schema where the page clearly describes an offered product

• FAQPage schema on real FAQ sections

• Article schema on blog content

• HowTo schema on true step-by-step tutorial content

This is mechanical work. It doesn't replace anything else on this list. If you miss it, you leave structured signals on the table that crawlers and search systems can use, and it's an easy win for a developer who can spend an afternoon getting it right.

Step 6: track citation share, iterate, repeat

The reason most AEO programs stall is that nobody is tracking the right metric. Page views are not the metric. Keyword rankings are not the metric. Backlink count is not the metric.

The metric is citation share on your 30 to 50 high-intent questions across the five answer surfaces.

Run the baseline. Make the changes in steps 3, 4, and 5. Re-run the same questions 30 to 60 days later. Compare. Identify the questions where you moved, the questions where you didn't, and the surfaces where you're still invisible. Iterate.

This is a 90- to 180-day cycle. Anyone promising you cited mentions in three weeks is selling you something they can't deliver. The engines update their training and retrieval on their own cadences, and the third-party layer takes time to build.

Common mistakes founders make on AEO

Treating AEO like SEO with new words. Same content calendar, same blog cadence, same keyword research, with "AEO" stamped on the project. The playbook is different. Most of the work isn't on your domain.

Optimizing for the wrong metric. Page views and rankings are SEO metrics. The AEO metric is citation share. They don't move together, and a content program that improves the first one can do nothing for the second.

Writing more content instead of fixing the content you have. Most SaaS sites have plenty of pages. The pages are just shaped wrong for AEO. Rewriting beats writing.

Ignoring the third-party layer. This is where most of the lift is, and most teams skip it because it's harder to staff than blog content. If you only have time for one of the six steps, this is the one.

Hiring an SEO agency to do AEO. They will mostly do SEO. The instincts are different. The motion is different. Pick a partner whose primary work is on the third-party signal layer, not on your blog.

Trying to optimize all five answer surfaces at once with the same playbook. ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews pull from overlapping but distinct source sets. The fastest wins usually come from picking the two surfaces your category buyers actually use and prioritizing those.

Questions I get constantly

How long does AEO actually take to show results? Real lift takes 90 to 180 days. The engines update their training and retrieval on their own cycles, and citation that lands and holds requires the underlying signal layer to be present.

Can I do this in-house? Yes, if you have a senior content person who's already obsessed with how the engines work. Most teams don't. The cost of hiring that person is higher than working with a specialist for a year. The right answer for most B2B SaaS companies is "agency for the first 6 months, in-house after we know what the loop looks like."

Is this just for B2B SaaS? No. Any category where buyers ask "what's the best X for Y" is in the AEO game. B2B SaaS is the most obvious case because the buying cycle has high research surface area, but consumer DTC, professional services, and B2B services all play.

What's the difference between AEO, GEO, and SGE? Different acronyms for overlapping playbooks. Answer Engine Optimization, Generative Engine Optimization, and Google's older Search Generative Experience label. The terminology hasn't settled. The work is mostly the same.

Should I worry that Google still drives most of my traffic? Today, yes, mostly. Two years from now, no. The transition isn't 0 to 1. It's gradual erosion of traditional search clicks as buyers default to AI assistants for research. The brands that start optimizing now will be cited when the share flips, and the brands that wait will be trying to play catch-up against incumbents the model has already learned to recommend.

Do paid ads work in AI search? Yes, but not yet in the way they work on Google. ChatGPT is testing ads on Free and Go tiers, with ads separated from the organic answer. Claude says it will remain ad-free. Perplexity has tested sponsored follow-up questions instead of paid recommendations inside the core answer. The smart play is to assume paid placements will keep expanding and that your organic citation share is the moat that survives when ad formats move.

How to optimize for AEO from here

If you've been wondering how to optimize for AEO and most of what you've read sounded like the same SEO advice with chatbot vocabulary on top, that's because most of it was. The actual playbook is what's above. It's not glamorous. It's not a trick. It's six steps run on a 90-day loop.

Map the questions your buyers actually ask. Measure your baseline citation share. Rewrite your content for passage extraction. Build the third-party signal layer. Get the schema right. Track citation share and iterate.

The reason most teams are losing in AI search isn't that they couldn't figure this out. It's that nobody on their team had the explicit job of doing it, and the SEO function they already have is optimizing for a different problem entirely.

The agencies that get this right are doing exactly the above. So can your team. The mechanics aren't proprietary. The discipline of running the loop, week after week, is.