AEO playbook: getting cited by ChatGPT, Perplexity, Gemini

In Q1 2026 we crossed a threshold most operators haven't named yet. Across the four properties we operate, 22% of inbound traffic now arrives via an AI answer engine — ChatGPT Search, Perplexity, Google AI Overviews, Gemini, Claude — rather

In Q1 2026 we crossed a threshold most operators haven't named yet. Across the four properties we operate, 22% of inbound traffic now arrives via an AI answer engine — ChatGPT Search, Perplexity, Google AI Overviews, Gemini, Claude — rather than via a classic search-result click.

The fraction was 4% in Q1 2025. The doubling cadence is quarterly. Gartner's projection of a 25% drop in traditional search volume by end of 2026 is, in our data, conservative. The shift isn't theoretical anymore. It's already in the analytics.

The question that's now load-bearing for every content-driven business: when an AI assistant answers a query in your category, does it cite you? Answer engine optimization in 2026 is the discipline of structuring content so that the LLMs actually pulling answers from the web pick your page as a source — and it's the work most operators are skipping because their dashboards still measure clicks, not citations. This is the six-lever playbook, the measured citation-rate curves, and the work that compounds versus the work that doesn't.

The AEO landscape has stabilized — but the work hasn't

Three things became true in the last twelve months that make AEO a real discipline, not a buzzword.

The answer engines themselves matured. ChatGPT Search exited beta in late 2025 and now reaches 883 million monthly users. Perplexity passed 25 million weekly actives. Google AI Overviews now appear on roughly 55% of all Google search results. Claude's web-search integration shipped in early 2026. Gemini's answer surface unified with Google's search backend. The five major engines all index the open web on a similar cadence — typically 24–72 hours from publication to indexing.

The citation mechanics became measurable. Engines started exposing source citations transparently — Perplexity in the UI, ChatGPT in the "sources" surface, Google in the AI Overview footnote. This means an operator can actually count which engine cited which page, when. Three years ago this was opaque; today it's a dashboard metric.

The traffic became real. AI-referred traffic crossed 5–25% of total organic traffic for most content-driven businesses in late 2025; by mid-2026 the band has widened to 8–35%. The high end is sites that have been doing AEO work intentionally; the low end is sites that haven't. The gap is the work.

The discipline now exists. The hard part is that almost nobody is doing the work yet — 70% of operators report AEO is strategically critical, only 20% have started.

The six AEO levers, ordered by what actually compounds

After running AEO work across the four properties for fourteen months, we've sorted the levers by what pays back. The order matters; the cheap-and-fast wins are at the top, the long-horizon work at the bottom.

Lever 1: answer-shaped content. Every page that wants to be cited should have the answer to its primary question in the first paragraph, in plain language, before any setup. The first 60–80 words should stand alone as an answer. The LLMs are looking for chunks they can extract; if the answer is buried at line 200, it doesn't get extracted.

Lever 2: structured data and schema markup. FAQPage, HowTo, Article schema with author + date + reviewedBy. The LLMs read the JSON-LD when deciding which page to cite from. A page with clean FAQPage schema is roughly 2–3x more likely to be cited from than the same content without it, in our measurement across ~150 pages we've tested with and without.

Lever 3: entity authority. Your company name, your products, your authors must appear consistently across the open web — Wikipedia (where possible), Crunchbase, LinkedIn company page, GitHub, your own About page, third-party reviews, industry directories. The LLMs use this entity graph to decide whether you're a real source or a thin one.

Lever 4: citation-friendly answer shapes. AI users ask compound questions: tradeoffs, comparisons, "best X for [my situation]." Single-keyword pages don't get cited as much as comparison/tradeoff pages. The same fact can be written as either; the comparison framing wins citations.

Lever 5: industry-platform presence. Reviews on G2, Capterra, Trustpilot, industry-specific platforms (Houzz for home services, Healthgrades for medical, Yelp for local) factor into how the LLMs read your authority. The engines read these aggregator pages and use them as cross-references.

Lever 6: freshness signals. Last-updated dates, recent comment activity, recent author additions. The LLMs prefer pages they can trust as current, especially for time-sensitive verticals. A 2024-dated post on a 2026 topic loses to a 2026-dated post even with weaker content.

Lever 1: answer-shaped content beats SEO-shaped content

The single highest-leverage structural change for AEO is moving the answer to the front of the page.

Classic SEO content patterns put context first — "in this article we'll explore," background, scene-setting — and the answer at line 200 or 400. This was always bad for human readers; the LLMs are now making it expensive in citation terms too. The LLMs are looking for extractable answer chunks in the first few paragraphs of a page. If the chunk isn't there, they cite a different page.

The pattern that works:

Paragraph 1: the direct answer to the page's primary question, in 60–80 words, in plain language. No setup, no "we'll cover," no "let's dig into."

Paragraph 2: the evidence — the specific number, the named source, the comparative claim that supports paragraph 1. This is the part the LLMs cite from, after they've decided paragraph 1 has the answer.

Paragraphs 3 onward: the nuance, the exceptions, the tradeoffs, the deeper context. This is what the human reader stays for. It's also what makes the page worth citing — a page that's only the answer-paragraph reads as thin; a page with answer + evidence + nuance reads as authoritative.

We've rewritten roughly 40 of our own corpus pages to this pattern over the last two quarters. Citation rate per page lifted 35–60% on the rewrites, with the biggest gains on comparison/tradeoff posts. Posts that already had answer-first structures (typically our case studies, which open with the result) saw the smallest deltas — they were already doing the work.

Lever 2: schema markup is the cheapest 2–3x you'll ever get

Schema markup is the part of AEO that gets dismissed because it's technical and unsexy. It's also the cheapest measurable lift available.

The three schema types that matter for AEO in 2026:

FAQPage — for any page with question-and-answer content. Faster citation than equivalent unstructured Q&A. We mirror our FAQ frontmatter into FAQPage JSON-LD on every post; the schema sits in the page head and the LLMs read it as a direct map of "this page contains answers to these specific questions."

Article — with author, datePublished, dateModified, publisher, and ideally reviewedBy. The author + date is what the LLMs use to decide if the page is fresh and authoritative.

HowTo — for procedural content. The LLMs love HowTo schema because it gives them ordered steps they can cite as a list.

The work to add schema is small. A @type: FAQPage JSON-LD block at the bottom of a page is 20 lines of code per template; we generate ours automatically from the post's frontmatter array. The lift is 2–3x citation rate vs the same content without schema, measured across ~150 pages we A/B tested over 8 weeks in Q1 2026.

Most AEO consultants we've audited skip this lever because it doesn't look like content work. Adding schema across an entire site is the single fastest measurable AEO lift we know of.

Lever 3: entity authority is the long-horizon work

The LLMs don't just read the page they're considering. They read the open web for context about who you are. This is the "entity graph" — your company name, your products, your authors, your domain, mapped against everything the LLMs have ever read about all of those.

The work to build entity authority is unglamorous and slow.

Consistency. Use the same company name spelling everywhere. The same author name on every byline. The same product name in every reference. We've audited brands where the same product appears as "Type A Bundle," "TypeA Bundle," and "Type-A Bundle" across LinkedIn, the website, and third-party reviews. The LLMs treat these as three different products, none of which has enough mentions to be authoritative.

Coverage. Wikipedia page if possible (small startups usually can't). Crunchbase listing. LinkedIn company page with active posts. GitHub organization. Industry-specific directory listings (G2 for SaaS, BBB for local services, Houzz for home services, Healthgrades for medical). Each of these adds an entity-graph signal.

Cross-references. Other people citing you. This is the part operators can't manually produce, but they can earn by being citation-worthy — by publishing original research, taking strong positions in podcasts, getting quoted in trade press. The third-party citations are what the LLMs use to weight your authority above peers.

Entity work is the part of AEO that takes 14–20 weeks to show in citation lift. We've watched two client sites cross the entity-authority threshold (when their company name starts appearing as a cited source in unprompted answers, not just in answers about themselves) — both took roughly six months from the start of intentional entity work.

This pattern overlaps directly with how LLM-cited blog content compounds: the work is cumulative, not transactional. Every cited page makes the next page slightly more likely to be cited.

Lever 4: write for compound questions, not single keywords

The shape of AI-user queries differs from the shape of Google-user queries, and the content that wins differs accordingly.

Google users type 2–4 word queries. "Best AI receptionist," "local SEO checklist," "WhatsApp business automation." The content that ranks for these is content that names the topic and answers a single direct question.

AI users ask compound questions. "What's the difference between AIRA and Smith.ai for a one-location dental practice in Munich?" "Is local SEO worth it if I'm only serving five postal codes?" "How does WhatsApp's 2026 policy affect a custom chatbot vs an off-the-shelf one?" The content that gets cited for these is content that compares, contextualizes, and names exceptions.

The same fact can be written either way. "AIRA costs $24.95/month" — single-keyword answer. "AIRA at $24.95/month beats Smith.ai's $99 tier for one-location operations under 100 calls; the math flips at 300+ calls or when the operation needs HIPAA compliance" — compound-question answer. The second version gets cited.

The structural change in our content is that we now write H2 headings as compound statements, not as questions. "When renting is the right answer" (compound — implies "vs the alternative"). "What still needs a human" (compound — implies "and what doesn't"). The H2 carries the comparison; the body carries the evidence.

This is also where the AEO playbook converges with the Suits-coded voice we've documented across other strategic-essay pieces — sharp comparisons land better with both human readers and LLM citation engines than dispassionate single-keyword pages.

Lever 5: industry platforms — where the LLMs cross-check authority

A pattern we noticed at the citation-engine level: when an LLM is about to cite a source, it cross-checks against industry-specific platforms before committing. A dental SaaS getting cited on a "best dental scheduling software" query gets cross-checked against G2, Capterra, and Software Advice. A local plumber gets cross-checked against Yelp, BBB, and Google Business Profile.

The implication: presence on the relevant industry platforms is an indirect AEO lever. You're not optimizing those platforms for traffic; you're optimizing them for the entity-authority cross-reference.

The platforms that matter by vertical, in our experience:

• SaaS / software: G2 (primary), Capterra, GetApp, TrustRadius
• Local services: Google Business Profile (still the strongest cross-reference), Yelp, BBB, industry-specific (Angi, Houzz, Thumbtack)
• Medical: Healthgrades, Zocdoc, Vitals, GBP
• Legal: Avvo, Martindale-Hubbell, GBP, state bar listings
• Consumer/DTC: Trustpilot, Sitejabber, the brand's own /reviews page
• B2B services: Clutch, GoodFirms, The Manifest

The work is twofold: claim the listing, then keep it accurate. Stale listings (wrong phone, wrong address, wrong service list) are worse than no listing because they introduce contradictions the LLMs read as signals of unreliability.

We've watched a single corrected G2 listing (right product description, current pricing, three updated reviews) lift Perplexity citation rate on a B2B SaaS client by roughly 40% over the following six weeks. The work cost a single afternoon. The lift compounded.

Lever 6: freshness signals — the small thing that pays back

The LLMs heavily weight publication date and modification date when deciding which source to cite for time-sensitive queries.

For evergreen content this doesn't matter. For anything tied to a year (best X 2026, AI tools for Y in 2026, comparison of A vs B updated [date]), freshness is a primary citation factor.

The work:

Update dateModified in your schema when you update the page meaningfully. Don't update it for typo fixes — the LLMs sometimes catch this and downweight pages that lie about freshness.

Update the visible date on the page. The LLMs read the rendered DOM, not just the schema, and a date that says "Updated May 2025" on a page claiming to be a 2026 guide gets caught.

Add a small "What changed in this update" section if the post is a major rewrite. This gives the LLM a chunk to cite that itself signals freshness.

For a topic-cluster strategy, refresh the top 10% of cluster pages quarterly. The cost is small (1–2 hours per page); the citation-lift compounds because the LLMs notice the cluster is actively maintained.

What we'd ship first if starting AEO this week

A concrete sequence for an operator with an existing content site, ordered by leverage.

Week 1: Add FAQPage and Article schema to your top 20 pages. This is the highest-leverage technical change available; it's also the change every AEO consultant will charge $5,000+ for.

Week 2: Rewrite the openings of your top 10 highest-traffic pages to the answer-first pattern (paragraph 1: direct answer; paragraph 2: evidence; paragraph 3+: nuance). Track citation rate per page before and after.

Week 3: Audit your entity-graph: company name spelling consistency, author bylines, LinkedIn company page activity, primary directory listings. Fix the inconsistencies.

Week 4: Pick the top three industry-specific platforms for your vertical and ensure your listing is accurate, current, and has 3+ recent reviews.

Weeks 5–8: Identify the five compound questions a buyer in your category would actually ask an AI assistant. Write one post that answers each, in the answer-shaped format, with schema. These five posts are the seed of your AEO compounding curve.

Weeks 9–10: First measurement. Pull citation rates from each major engine. Iterate on what worked.

By week 10, the lift should be measurable. By week 20, it should be compounding. By week 52, the AI-referred traffic share should have crossed 15% if the work was done seriously.

Classic SEO felt like fighting for a position in a list of ten. AEO feels like fighting to be quoted in a paragraph of three. The strategic shift is small. The work is mostly the same disciplines applied differently. The operators who think AEO is a different game are going to spend money on the wrong thing.

— our SEO lead, after the third client AEO sprint measured genuine citation lift instead of the usual nothing

The discipline isn't new. The format the discipline takes is. Operators who've done classic SEO well already have most of the foundations — they just need to add the structural patterns (answer-first, schema, compound-question framing) that match the new index. Operators who've done classic SEO poorly will find AEO unforgiving for the same reasons they found SEO unforgiving — the algorithms reward signal, and signal is hard to fake.