Engineer pages so AI assistants cite you, not just rank you

01 Put AEO under GTM, not under SEO, before you touch a page.

This is an ownership decision, and getting it wrong upstream poisons everything downstream. SEO competes for a ranked slot on a known query. AEO competes for inclusion and a brand mention inside a generated answer, which depends on how your claims, comparisons, and category framing show up across the public corpus the model trained on or retrieves from. Those inputs are positioning, comparative framing, named alternatives, and customer-language vocabulary. They are PMM artifacts. AEO is a GTM capability, not an SEO experiment

"In SEO, you compete for rankings. In AEO, you compete for mentions."

· Maja Voje, AEO: How to Make AI Recommend Your Product, 2026-04-17 · “AEO is a GTM…”

"AEO isn't a content thing or an SEO experiment. It's a GTM capability."

· Maja Voje, AEO: How to Make AI Recommend Your Product, 2026-04-17 · “AEO is a GTM…”

If SEO or growth eats AEO first, the work optimizes for chunker-friendly text and misses the positioning and narrative levers that drive mention rate. Split it cleanly. PMM owns positioning, mention-rate strategy, and the comparison roadmap. Technical SEO owns passage engineering and structured data. The risk is letting one function eat the other. The AEO triangle, presence, relevance, manual-action propagation

02 Read the substrate and extract canonical entities first.

Before you write a passage, read the positioning, the ICP-anchor, the brand-voice rules, and the product knowledge for any product entities in the target artifact. Then extract the canonical entities: product names, integrations, competitors, ICP role titles, pricing tiers, capability claims. Output an entities.json with type tags compatible with schema.org Product, Service, and Organization.

Treat the entity list as a gate, not a step. Every entity and capability claim must trace to a substrate path. An entity that is not present in product knowledge or the competitive data bank does not go in. This is where AEO fabrication starts: a model latches onto a thin claim and the surface confirms it. Mike King's relevance-engineering frame is the reason entities come first: LLM retrieval scores chunks against entities and relations, not keyword density, so the entity graph is the substrate the rest of the pass engineers around. The unit of optimization is the passage, not the page

03 Clear the experience and structure gate before any content work.

Content investment is downstream of passing the performance, structural, and experiential gates. Run the infrastructure audit first, or you optimize copy on a page no crawler ever finished reading.

Check server logs for 499 responses filtered to AI bot user-agents (GPTBot, PerplexityBot, ClaudeBot). A 499 is a client-initiated close: the bot gave up. Once a page is skipped it never enters the candidate pool the model considers when generating a citation. AI crawlers return 499 errors on slow pages. Speed is a gate, not a ranking modifier.

"A 499 doesn't mean your server failed. It means the system requesting your content decided it wasn't worth waiting for."

· Mike King, Page Speed Impacts, 2026-05-08 · “AI crawlers return 499…”

Set TTFB and payload budgets per template, not site-wide. The crawl is only the first gate. Every major AI search platform now routes queries through five sequential gatekeepers: Planner, Retriever, Reader, Critic, Synthesizer. The Critic drops most content with no signal to the publisher, so content can appear in a Retriever-stage source panel and still vanish from the final answer. Single-shot retrieval optimization is obsolete against this. AI search runs a five-gatekeeper pipeline and the Critic is where most content dies, invisibly

"You cannot optimize what you cannot observe. Distilling your own version of it is the only path to durable GEO performance."

· Mike King, Agentic RAG and the Future of GEO, 2026-05-20 · “AI search runs a…”

Then audit the structural chrome. What you place in headers, subheaders, top nav, and footers is read as a relevance signal independent of body content. Nav and footer appear on every page, so their contents act as persistent topic declarations. Adding a use-case anchor to the footer is a site-wide statement of competence in that area. Nav and footer placement is a first-class LLM relevance signal, independent of page content.

"structural prominence or what you put in your headers, subheaders, top nav, and footer, matters for LLM citations"

· Casey Hill, What's Working Right Now in AI Search, 2026-05-03 · “Nav and footer placement…”

The content work is the last step, not the first. AEO experience and structure gate

04 Produce 8 to 12 stand-alone passages, 40 to 90 words each.

Each passage answers exactly one buyer question and stands alone, because LLMs cite passages, not pages. AI search runs a retrieval pipeline: chunk the doc into passages, embed each, score embeddings against a query fan-out where the model expands one query into many sub-queries, then assemble matching passages into a synthesized answer. A page that ranks first for a head term can still be invisible if its passages are too short, too long, or thin on entity density. The unit of optimization is the passage, not the page

"Answers are generated, not linked."

· Mike King, How AI Mode Works, 2026-04-22 · “The unit of optimization…”

Shape the passages as argument, not coverage. The ultimate-guide play loses its mechanic when an LLM is the summarizer: the model paraphrases facts brand-free, but it cannot compress a credible perspective without naming the perspective-holder. POV survives summarization where coverage gets dropped. Sharper POV beats exhaustive coverage when an LLM is the summarizer

"The ultimate-guide play loses its mechanic when an LLM summarizes for the user, so sharper POV beats exhaustive coverage."

· Amanda Natividad, The Death of the Ultimate Guide, 2026-04-25 · “Sharper POV beats exhaustive…”

Lead the comparison passages, not the how-to passages. Across 454 prompt-domain pairs and four AI engines, 62% of citations never name the brand. Comparative content (X vs Y, alternatives-to teardowns, ranked listings) forces the model to keep brand tokens together with claims, so the brand survives summarization where informational content gets summarized brand-free. Citation rate and mention rate are different metrics; comparative content closes the gap

"Being cited means an AI is drawing on your content. Being mentioned means it is naming you."

· Kevin Indig, The Ghost Citation Problem, 2026-04-26 · “Citation rate and mention…”

Add at least one primary-research or expert-perspective passage. High-reasoning AI mode fires far more queries per answer and cites a largely different set of domains, surfacing primary sources and structured evidence that keyword-optimized content does not earn. The overlap between minimal and high-reasoning citation sets is 25.6%, so a passage library built for one mode misses most of what the other cites. High-reasoning AI mode and standard mode share only 25.6% of cited domains; AEO built for one misses three in four sources the other cites

"The brand that wins under minimal reasoning is not the brand that wins under high reasoning."

· Kevin Indig, Reasoning Lift, 2026-05-18 · “High-reasoning AI mode and…”

05 Generate validated FAQPage and HowTo schema for the passages where the shape fits.

For passages with a clean question-and-answer shape, emit FAQPage and HowTo schema blocks. Validate every block against the current schema.org spec. Schema generated without validation is a failure mode, not a deliverable. The schema is the structured-data layer technical SEO owns, and it is what lets the entity graph from step 2 surface as machine-readable claims rather than prose a chunker has to infer.

06 Map each passage to a third-party mirror, because AEO is decided off your domain.

On-page work is necessary and not sufficient. Assistants pull citations from YouTube, Reddit, LinkedIn, and podcasts more than from a brand's own site, so the missing surface is third-party mention density across those venues. A brand mentioned across YouTube, Reddit, and LinkedIn long-form gains citation authority an on-domain push cannot replicate. AEO is decided off your domain, third-party mention density is the missing surface

The off-site weight is structural, not incidental. AI search visibility is an off-site corroboration problem with an on-site quality floor, not the inverse, so the diagnostic job before any on-site sprint is mapping which third-party surfaces already earn citations in your category. AI search visibility is an off-site corroboration problem with an on-site quality floor, not the inverse Output a mirror plan: per passage, name the venue, the title, the opening hook, and the source passage. Pull the demand language from sales, success, and support transcripts, not from keyword tools, which lag reality. AEO is decided off your domain, third-party mention density is the missing surface

07 Keep self-promotion out of the corroboration layer, or one penalty cascades everywhere.

The off-site move backfires if you run it as self-referential listicle farms or artificial timestamp refreshes. AI surfaces retrieve content via RAG from Google's index, so a Google enforcement action does not stop at Google search. It removes the page from the shared source pool that ChatGPT, Perplexity, and Copilot all draw from. One enforcement decision becomes a multi-surface traffic loss. A Google penalty removes a page from every AI surface at once, not just Google search

That's getting recommended by everybody else without recommending yourself. And to me, that's where you want to go.

· Lily Ray, What the SEO Industry Is Getting Dangerously Wrong About AI Search, 2026-05-13

A Google penalty removes a page from every AI surface at once, not just Google search

Run a brand-integrity watch in parallel. AI synthesis surfaces have a thin verification layer for sparsely-covered claims, so a single seeded fake claim can self-confirm by citing the source that introduced it. Brand protection now means weekly sweeps for hallucinated claims about your brand, measured separately from search-results monitoring. A single seeded fake claim can self-confirm in AI Overviews The AEO triangle, presence, relevance, manual-action propagation

08 Score on Aleyda's three layers and keep the confidence buckets separate.

Replace rank-tracking with a three-layer measurement stack, and never collapse it into a single AEO number. The value is in the separation. Measure AI search on three layers: Presence, Readiness, Business Impact

"Presence tells you where the brand appears, Readiness tells you why it looks that way, and Business Impact tells you whether that visibility creates measurable value."

· Aleyda Solis, A 3-Layer Framework to Measure AI Presence, Readiness, and Business Impact, 2026-04-23 · “Measure AI search on…”

Track recommendation rate, not just citation count, on the Presence layer. AI responses recommend specific products rather than listing links, so the competitive frame is owning the recommendation in a category query, not mining citations. The AEO metric is recommendation-share, not citation volume. AI responses recommend products; they do not list links.

"AI responses recommend products rather than simply provide a list of links"

· Kyle Poyar, What's Working Right Now in AI Search, 2026-05-03 · “The AEO metric is…”

On the Business Impact layer, hold the line between citations and traffic. More AI answer links do not automatically mean more traffic: a citation can appear without a linked URL, a link without a click, a click without a conversion, and each break needs a different fix. More AI answer links do not automatically mean more traffic; track six AEO signals separately, not blended into one score

"More AI answer links don't automatically mean more traffic."

· Aleyda Solis, SEOFOMO, May 10, 2026 · “More AI answer links…”

09 Measure weekly with correlated short signals, not point-in-time snapshots.

A single-snapshot citation count is noise. Roughly 74% of cited sources in AI search rotate week to week, so a monthly reporting cycle samples a different distribution each time and every decision rests on data already stale. Point-in-time AEO citation counts are noise: 74 percent of cited sources rotate weekly A weekly, segmented measurement layer is the minimum structure that separates signal from variance: track by platform, prompt type, citation count, recommendation rate, readiness, and business impact, separately. Absolute counts + correlated short signals, not stage rates and long loops

The underlying construct is dissolving too. Keyword position reporting measures an aggregate that loses its referent as Google personalizes results per user. There is no longer a single stable position to rank for, which is the structural reason the three-layer stack and weekly sweeps beat a position dashboard. Keyword position reporting is measuring a construct that is dissolving as Google personalizes results per user Pin the goal to correlated short signals you can read weekly, not a long-loop number you read once a quarter. Absolute counts + correlated short signals, not stage rates and long loops