AEO vs SEO: what changed in 2026 and what to do about it

The SEO playbook stayed mostly stable from 2010 to 2023. Then in 18 months it didn't.

By Q1 2026, AI-driven search (ChatGPT, Claude, Perplexity, Google AI Overviews, Copilot, and the dozen smaller engines that have arrived since) accounts for 15–30% of B2B inbound traffic for most sites we audit. AI-referral traffic also converts at 4–5× the rate of classic organic search. That's not a margin to ignore.

This piece is the practical version of "what's actually different about SEO in 2026." It's deliberately concrete. The goal is that you can read it, audit your own site, and ship the highest-leverage fixes this week.

#The big change in one sentence

Classic SEO optimizes for Google's ranking algorithm picking your link out of ten. AEO optimizes for an AI engine picking your content out of dozens to cite inside its answer.

The substrate is different. The classic Google result is a list of links: the user picks. The AI engine result is a single synthesized answer with citations: the engine picks. That changes everything downstream.

#What carries over from classic SEO

Most of it. About 70% of the AEO playbook is just well-executed SEO. Don't throw out what you know.

• Authoritative long-form content still wins. AI engines preferentially cite long-form (≥1,500 word) authoritative pages over thin pages, just like Google.
• Page speed matters. AI crawlers timeout faster than Google's; a slow page is a skipped page.
• Internal linking still propagates topical authority within your site.
• Backlinks still matter, but the math is different. High-quality co-citations from authoritative sources matter much more than backlink volume in 2026.
• Search intent matching still matters. Comparison-intent queries ("X vs Y"), pricing-intent queries ("how much does X cost"), and how-to queries are highest-citation surface areas.

If your classic SEO is bad, your AEO will be bad. Fix the SEO basics first.

#What's genuinely new in 2026

The five things that didn't matter in classic SEO and matter a lot in AEO.

#1. JSON-LD schema is no longer optional

In classic SEO, schema markup was a "nice to have" that influenced rich-snippet eligibility. In AEO, AI crawlers parse JSON-LD an order of magnitude faster than unstructured HTML, so pages without schema get evaluated more shallowly or skipped entirely.

The minimum schema for any B2B SaaS page in 2026:

• Organization on every page (founder/about info, contact, social links)
• WebSite on the home page (with SearchAction if you have site search)
• BreadcrumbList on every non-home page
• FAQPage on every page with explicit Q&A
• Service + OfferCatalog on every service or product page
• Article + Person (author) on every blog post and case study
• HowTo on any process / tutorial page

This is not optional in 2026. The cost of adding it is one engineer-day per page type. The cost of not adding it is being skipped by every AI engine for citations.

#2. FAQ-structured content gets cited disproportionately

Conversational AI is built around question-answering. Pages structured as explicit FAQPage Q&A blocks are the highest-citation surface in AEO. They're literally the format the AI is built to consume.

What this means in practice:

• Every service page should have a 5–10 question FAQ block
• Every blog post should have a 3–5 question FAQ block at the bottom
• The FAQs should answer the questions your ICP is actually asking ChatGPT. Not the questions your marketing team wrote in a brainstorm

How to find the right FAQs: open ChatGPT, Claude, and Perplexity. Type 10–20 prompts your ICP would type ("how much does it cost to build a SaaS MVP", "best AI automation agency for B2B", "n8n vs Zapier"). Read the answers. The structural questions in those answers are your FAQ list.

#3. llms.txt and llms-full.txt are real signals now

llms.txt is an emerging web standard, like robots.txt for AI crawlers. It lives at yourdomain.com/llms.txt and gives AI crawlers a curated index of your most important pages with one-line summaries. llms-full.txt lives alongside and contains the full markdown of every canonical page concatenated.

Adoption in early 2026: not universal but growing. ChatGPT's web crawler, Perplexity, and several smaller engines explicitly read it. Google's AI Overviews crawler does not yet, but the AI-search community treats llms.txt as a positive signal regardless.

The cost of providing it is essentially zero. It generates from the same content that drives your sitemap. There is no reason not to ship it.

#4. Author identity (E-E-A-T) is load-bearing

Google's E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) framework was introduced for medical and financial content. In 2026, AI engines apply E-E-A-T-style filtering to all content, not just YMYL niches.

Specifically: AI engines preferentially cite content that has a real, named author with a well-connected sameAs graph (LinkedIn, GitHub, X, prior published work). A blog post by "the team" cites less reliably than a blog post by a named senior engineer with a public profile.

This is a deliberate tradeoff for some sites. If you can't or won't name authors publicly, you'll lose some AEO citation potential. The signal still works at the Organization level, but it's noticeably weaker.

If you can name authors:

• Real name, real photo, 250+ word bio on an About / author page
• Person schema on every article with full sameAs array
• Visible "Last reviewed" + "Published" dates that match dateModified schema
• First-person voice in long-form (write "I think", not "the team thinks")

If you can't (privacy, anonymity, internal policy):

• Lean harder on Organization schema with rich sameAs
• Make the org's distinct expertise visible through content depth, not personal identity

#5. AI crawler accessibility. The underrated AEO signal

A surprising number of sites still block AI crawlers in robots.txt. Usually it's because someone added it during the 2023 panic about "AI stealing our content." It's almost always the wrong call in 2026.

The crawlers to explicitly allow in 2026:

User-agent: GPTBot
Allow: /

User-agent: ChatGPT-User
Allow: /

User-agent: OAI-SearchBot
Allow: /

User-agent: ClaudeBot
Allow: /

User-agent: anthropic-ai
Allow: /

User-agent: PerplexityBot
Allow: /

User-agent: Google-Extended
Allow: /

User-agent: CCBot
Allow: /

User-agent: Applebot-Extended
Allow: /

User-agent: MistralAI-User
Allow: /

These are the bots that actually read your content for AI search and AI training. Blocking them is invisibility. Allowing them costs you nothing. They're not displacing your traffic, they're redirecting it via citations.

There's a separate question of "should I block AI training" (the crawlers that gather data for model training, not for live search). That's a values call. But the search crawlers should always be allowed.

#The 2026 AEO checklist

If you do nothing else this quarter, do these five things in order:

1. Allow all AI search crawlers in robots.txt. One-line edit. Ship today.
2. Add JSON-LD schema to every page type. Organization on every page; Service/Article/FAQPage where appropriate. One engineer-day per page type.
3. Ship llms.txt at yourdomain.com/llms.txt. Should be a single index file with links to every canonical page and a one-line summary each.
4. Add explicit FAQ blocks to every service page and pillar blog post. Using FAQPage schema. The questions should be the ones your ICP is actually asking AI engines, not your marketing team's brainstorm.
5. Establish author identity (or commit to brand-only voice). If you're naming authors, ship Person schema on every article with full sameAs to LinkedIn / GitHub / X / prior published work. If not, lean hard on Organization E-E-A-T signals.

These five fixes will move citation-share noticeably within 4–8 weeks. Everything beyond this is optimization on top.

#What to measure

Classic SEO metrics (rank position, organic traffic) still matter but they're not enough in 2026. Add these to your dashboard:

• Citation share by query. For 10–20 prompts your ICP would actually type, run them monthly through ChatGPT, Claude, Perplexity, and Gemini. Log whether you're cited, mentioned, or absent. Track citation share against named competitors.
• AI-referral traffic. Measurable in GA4 by referrer (look for chat.openai.com, perplexity.ai, gemini.google.com, claude.ai, etc.)
• Conversion rate of AI traffic vs organic. Should be 4–5× organic in 2026; if it's not, your site is failing to qualify the AI-referred user (usually a hero clarity problem)
• Schema validity score. Every page validating in Google's Rich Results Test and the schema.org validator

Tools that help in 2026: Athena and Profound for citation tracking, Ahrefs and Semrush still for classic SEO, Google Search Console + Bing Webmaster Tools for both classic and AI-Overview-style traffic where measurable.

#The strategic shift

Beyond the tactics: AEO requires a different kind of content than classic SEO. Classic SEO rewarded broad topical coverage and link-bait. AEO rewards narrow topical authority and primary-source depth.

What this means in practice:

• Fewer, deeper pieces. A 5,000-word definitive guide on one topic outperforms ten 800-word posts on related topics.
• Original analysis over aggregation. AI engines actively de-prioritize content that reads as derivative or AI-generated; first-person experience and primary data win.
• Specific answers over hedge-y prose. "It depends" content gets cited less than content that says "in 2026, n8n is the right call for X / Make for Y / Zapier for Z."
• Frequent updates over set-and-forget. Pages with dateModified fresher than 90 days get cited more frequently than pages last updated 18 months ago. The Princeton GEO study confirms that statistics-rich, citation-dense, recently-updated content drives 30–40% visibility lift across AI engines.

#When AEO efforts are wasted

Three patterns we see waste AEO budget:

1. AI-generated content at scale. Counter-intuitively, AI engines actively de-prioritize content that reads as AI-written. Hand-written long-form with first-person voice and primary-source citations is the entire game.
2. Programmatic SEO templates. Pages built from a database template and minimal hand-written content. They were briefly effective in classic SEO; they're invisible in AEO.
3. Schema injection without underlying content quality. Adding FAQPage schema to a thin page with weak FAQs doesn't help. The schema works because it makes good content easier to parse. It doesn't manufacture authority.

#The next 12 months

Three things we expect to change between now and Q2 2027:

1. llms.txt becomes a de-facto standard. Adoption goes from "growing" to "expected." Sites without it become slightly invisible.
2. AI Overviews and ChatGPT Search become the default first stop for B2B research queries. Citation share starts to matter more than ranking position for some categories.
3. Author identity vs anonymity becomes a real strategic choice. Some brands will lean hard into named-operator visibility (huge AEO upside); others will go pure brand-voice (privacy upside, AEO cost).

The teams that win are not the ones that game any single signal. They're the ones that ship genuinely authoritative, well-structured, fast, AI-readable content. And let the citation share compound month over month.

#Where to start

If you're auditing your own site, the order is: robots.txt allow → schema baseline → llms.txt → FAQ blocks → author/E-E-A-T. Ship those five and you've covered 80% of the technical AEO work.

The remaining 20% is content quality, and that's the part that takes 6 months to compound regardless of how fast you ship the technical infrastructure.

If you want help auditing the technical baseline, that's literally what our AEO Audit is. A one-week productized engagement that ends with a 30-day fix plan and baseline citation tracking. Or you can implement everything in this post yourself; the playbook is intentionally written to be self-serve.