Introduction
Most AEO advice you read in 2026 treats AI search as a single decision: did the model cite your page or didn't it? That framing isn't useful when you're trying to fix a page that isn't getting cited, because there's no single point of failure. AI search is a multi-stage pipeline, and a page can fail at any one of those stages for entirely different reasons. A page that never gets crawled looks the same in your dashboard as a page that gets crawled, indexed, and rendered but loses on framing — the citation slot is empty either way.
That's the argument behind a useful diagnostic published on Search Engine Land by Jason Barnard on May 5, 2026. Barnard breaks the AI search journey into ten distinct gates a piece of content must pass to actually appear in a generated answer. Five are infrastructure gates — discovery, selection, crawl, render, index — and five are competitive gates — annotation, recruitment, grounding, display, and the final “won” gate where the brand narrative actually survives extraction.
The framework is dense and written for senior SEO practitioners. This guide translates it into a diagnostic flow any small business can run on its own site, gate by gate, with a checklist of what to inspect at each stage and which fix the failure suggests. We'll close with a 10-question self-audit you can complete in an afternoon. The goal isn't to memorize the model — it's to know which gate is actually breaking when a page isn't surfacing.
Key Takeaways
- AI search is a 10-stage pipeline, not a single ranking decision; content can fail at any stage for entirely different reasons
- The first five gates are infrastructure (discovery → render → index); the last five are competitive (annotation → display → won narrative)
- In a multiplicative pipeline, the weakest stage limits everything downstream — Barnard calls this “better to be a straight C student than three As and an F”
- Most small business pages fail at gates 4 (render) or 6 (annotation) — both are fixable in days, not months
- Diagnosing the failing gate is faster and cheaper than rewriting the page
- The 10-question self-audit at the end of this guide can be run in an afternoon
Why a Pipeline Model Matters Now
The traditional SEO audit checks rankings and assumes one thing produced them. AI search breaks that model because AI systems run multiple sequential evaluations before generating an answer. As Barnard frames it, each gate is multiplicative — a near-miss at one gate can zero out perfect performance at every other gate. He attributes the framing to Brent D. Payne and Gary Illyes' work on Google's ranking model: “better to be a straight C student than three As and an F.”
That logic matters because it changes how you triage a page that isn't surfacing. The instinct most operators have is to rewrite the content. That's the right move when the failure is at the competitive gates — annotation, recruitment, grounding, display, won. It's the wrong move when the failure is at the infrastructure gates — a page that doesn't get crawled doesn't need a content rewrite, it needs a sitemap fix. We covered the underlying structural-vs-distinctiveness split in our earlier post on why topical authority alone isn't enough, and the pipeline model is a practical extension of the same point.
The honest tradeoff with a 10-gate model is complexity. You don't need to memorize every gate to use the framework. You need to be able to ask, in order: did the page get found, get fetched, get understood, get selected, get framed correctly? Each of those questions maps to two gates, and the diagnostic is a binary at each one — either it passed or it didn't. A page that's stuck at gate 3 doesn't need gate 9 work yet.
What Are the Five Infrastructure Gates?
The first five gates are technical. They determine whether your content is even available to be considered. Every AI search system runs through approximately the same sequence — Google AI Overviews, ChatGPT search, Perplexity, Claude's web search, Gemini — though each weights the signals slightly differently.
Gate 1 — Discovered. Bots find URLs through sitemaps, IndexNow submissions, internal links, and inbound links. Failure mode: the URL exists but no crawler has been told it exists. Diagnostic: pull your XML sitemap, confirm the URL is in it, and check Google Search Console's URL Inspection for “discovered” status. If a URL isn't discovered, no other gate matters.
Gate 2 — Selected. Bots have a budget. Among discovered URLs, they choose which to actually fetch based on link prominence, anchor text, and historical signal quality. Failure mode: page is in the sitemap but never crawled. Diagnostic: check server logs for crawler activity on that URL over the last 30 days. We've covered the methodology in detail in our guide on log file analysis for AI crawlers.
Gate 3 — Crawled. The fetch attempt actually completes successfully. Failure mode: redirects, rate limits, server errors, robots.txt blocks, or 5xx responses. Diagnostic: GSC's Page Indexing report and server log status codes. A 200 response is the floor, not the ceiling.
Gate 4 — Rendered. The system can process the content it fetched. This is where JavaScript-heavy pages most often break — many AI crawlers don't execute JavaScript at all. Failure mode: critical content is inserted by client-side JS and the bot sees an empty or shell page. Diagnostic: fetch the URL with curl or “View Source” and check whether your headlines, body, and structured data are present in the raw HTML. We wrote about the fix in no-JavaScript fallbacks for crawlers — most modern frameworks support server-rendered fallbacks but require explicit configuration.
Gate 5 — Indexed. The rendered content gets stored in a way the system can later retrieve. Failure mode: page is rendered but excluded from the index due to canonicalization issues, duplicate content, soft 404s, or content quality flags. Diagnostic: GSC Page Indexing reports labeled “Crawled — currently not indexed” or “Discovered — currently not indexed.” These are the diagnostic signals to act on.
The infrastructure gates fail invisibly. Most operators don't realize the page never made it past gate 3 until they look. The good news is they're cheap to fix once identified — none of the failures above require rewriting the page.
What Are the Five Competitive Gates?
Once the infrastructure gates clear, the page enters the competitive layer. This is where most AEO writing focuses, and where most paid tools try to measure. Here the question is no longer “is the page available?” but “is the system choosing it over alternatives?”
Gate 6 — Annotated. The system attaches metadata: which entities are on the page, which topics, which schema types, which authorship signals. Failure mode: missing or thin schema markup, ambiguous entity references, low-confidence topic classification. Diagnostic: run the page through Google's Rich Results Test and Schema Markup Validator. If the system can't tell whether your page is about a service, a location, an article, or a product, it can't slot it into the right answer pool.
Gate 7 — Recruited. The page gets pulled into the candidate set for relevant queries. Barnard frames this as the “Algorithmic Trinity” — recency, originality, and clarity together determine whether the page makes it into the candidate pool a generator selects from. Failure mode: page is technically indexed and annotated but doesn't appear in any candidate set because the system reads it as derivative or dated. Diagnostic: check whether the page surfaces in Search Console for any conversational long-tail queries; if not, recruitment is failing.
Gate 8 — Grounded. The system uses the page to ground a claim — meaning it reads the page as proof for a specific assertion. Failure mode: claims are present but the page lacks authoritative entity identity, proof links, or N-E-E-A-T-T signals (a 2026 evolution of E-E-A-T that adds notability and trust factors per Barnard's framework). Diagnostic: ask three different AI assistants the question your page is meant to answer, and inspect whether your page is among the cited sources. If competitors are being cited and you're not, grounding is the failed gate.
Gate 9 — Displayed. The page survives the final filter where the generator decides which sources actually appear in the surfaced answer. Failure mode: the page was a candidate and was used internally, but the framing didn't match what the user-facing answer required. Diagnostic: this is the hardest gate to inspect directly; it's usually inferred from grounding-with-no-display patterns.
Gate 10 — Won. The brand narrative survives extraction. The system uses your content but rewrites titles, descriptions, or context in a way that strips your framing. Failure mode: you get a citation that doesn't actually serve your business — wrong context, missed nuance, attribution to the wrong service. Diagnostic: read the actual generated answers and check whether the framing aligns with how you'd want a prospective customer to encounter your business.
The competitive gates can usually be improved with content work, but the order matters: gate 6 (annotation) is the cheapest to fix and unlocks gates 7–10. Most pages we audit fail at gate 6, then at gate 8.
Where Most Small Business Pages Actually Fail
In our experience auditing small business sites across Northeast Indiana, the failure distribution is heavily front-loaded. Here's the rough shape we see.
| Gate | Approximate share of failures we see | Typical fix cost |
|---|---|---|
| 1–3 (discovery, selection, crawl) | ~15% | Hours to a day |
| 4 (render) | ~25% | Days, mostly framework configuration |
| 5 (index) | ~10% | Hours to a few days |
| 6 (annotation / schema) | ~25% | A focused week of schema and entity work |
| 7–8 (recruitment, grounding) | ~20% | One to three months of content distinctiveness work |
| 9–10 (display, won) | ~5% | Ongoing narrative discipline |
The implication is that more than half of all small business AI search failures are infrastructure or annotation problems — fixable in days, not quarters. That's good news. The bad news is that most small businesses we talk to are spending content effort on gates 7–10 before fixing gates 4–6, and the multiplicative logic of the pipeline means they get nothing for the work.
The shape also helps explain why a Search Engine Land analysis of why content fails in AI Overviews found, citing BrightEdge data, that the overlap between top-10 organic results and AI citations rose from 32.3% to 54.5% between May 2024 and September 2025 — but “even at peak convergence, nearly half of all AI Overview citations come from pages that don't rank at the top.” Traditional ranking helps, but it only addresses some of the gates. A page can rank well organically and still fail at gate 4 (render) for AI-specific crawlers, or fail at gate 8 (grounding) because it lacks distinct evidence.
How Do You Diagnose Which Gate Is Failing?
Here's a 10-question self-audit you can run on a single page in roughly two hours. Each question maps to one gate. Answer each one yes or no honestly; the first “no” is your starting point.
- Discovered: Is the page present in your XML sitemap and reported as “discovered” in Google Search Console's URL Inspection?
- Selected: Has any crawler — Googlebot, GPTBot, ClaudeBot, PerplexityBot — actually fetched the page in the last 30 days according to your server logs?
- Crawled: Does the URL return a clean 200 status with no redirect chains and no robots.txt block?
- Rendered: When you fetch the URL with
curland “View Source,” do all your headings, body text, and JSON-LD schema appear in the raw HTML — without requiring JavaScript execution? - Indexed: Does the GSC URL Inspection tool report the page as “URL is on Google” with no “Crawled — not indexed” warning?
- Annotated: Does Google's Rich Results Test parse your structured data without errors, and does the schema accurately label the page's primary entity (article, service, product, FAQ, location)?
- Recruited: When you check Search Console for the page over the last 30 days, does it appear with impressions on conversational long-tail queries (5+ words, often question-form)?
- Grounded: When you ask ChatGPT, Perplexity, and Google AI Mode three real questions the page is designed to answer, is your page among the cited sources at least once?
- Displayed: When the page is cited, does the generator actually surface its framing in the user-facing answer, or is it cited internally without showing in the response?
- Won: When the framing surfaces, does it represent your business accurately — service categorization, geography, value proposition?
Stop at the first “no.” That's the gate to fix. Don't move past it until it clears, because the multiplicative logic means downstream improvements won't compound until the upstream gate passes.
What Does Each Gate Failure Look Like — and What Fix Does It Suggest?
A short table mapping each failed gate to the practical next move.
| Failed Gate | Common cause | Where to start |
|---|---|---|
| 1 Discovered | Missing or stale sitemap; orphaned page | Regenerate sitemap; add internal links from indexed pages |
| 2 Selected | Low link prominence; weak anchor text | Add internal links with descriptive anchors; check inbound links |
| 3 Crawled | 4xx/5xx errors; robots.txt block; rate limits | Fix server errors; review robots.txt for AI crawler agents |
| 4 Rendered | JavaScript-only content; missing SSR | Add server-side rendering or static fallbacks |
| 5 Indexed | Duplicate/canonical issues; soft 404; thin content | Resolve canonicals; expand content quality |
| 6 Annotated | Missing or invalid schema; entity ambiguity | Add JSON-LD per Schema.org; clarify primary entity |
| 7 Recruited | Generic content; no recency; weak topic clarity | Add freshness signals; sharpen topical focus |
| 8 Grounded | No distinctive evidence; weak proof links | Add proprietary data; cite first-party sources |
| 9 Displayed | Framing doesn't match query intent | Restructure with question-format headings |
| 10 Won | Brand narrative gets stripped in extraction | Tighten consistent brand entity language site-wide |
For gates 7–10, the underlying lever is distinctiveness — the same lever we covered in information gain audits. Most pages that pass gates 1–6 still fail at gates 7–10 because they say what every other source says. The fix is contribution, not optimization.
For gate 8 in particular, the practical advice from a recent analysis of how AI models understand brands by Jes Scholz on Search Engine Land is to “reduce entropy” through consistency — canonical brand bio, consistent naming, repeated key associations across platforms. AI grounding rewards predictable identity, not variety. We've also written about this from the citation angle in what makes ChatGPT cite you, where the data showed citations correlate with traditional ranking and tight content focus.
Northeast Indiana Application: A Service Business Walkthrough
Most of the small businesses we work with in Fort Wayne, Auburn, and Allen County have more headroom in gates 4–6 than they realize. A typical pattern: a Fort Wayne home services business has a clean WordPress install with reasonable content. Gate 1 passes (sitemap is fine). Gate 2 passes (Googlebot is crawling). Gate 3 passes (no server errors). Gate 4 fails — half the body content is loaded by a marketing widget that fires after page load, and AI crawlers see a near-empty page. The site is invisible in AI search not because the content is weak but because the widget configuration is hiding it.
A second pattern: a multi-location dental group with strong location pages, all server-rendered, gates 1–5 clean. Gate 6 fails — the location schema is incomplete, listing the practice as a generic LocalBusiness instead of Dentist, with no medicalSpecialty properties. The annotation gate fails silently, and AI Overviews skip the practice when answering “Auburn pediatric dentist” because it can't tell from the schema that pediatric dentistry is offered. Fix the schema, and the same content suddenly starts surfacing.
These aren't hypothetical. They're the most common diagnostic outcomes we see, and both are gate 4 or gate 6 failures that no amount of content rewriting would have fixed. The pipeline framework matters because it forces the diagnosis to happen before the work, not after.
The growth in AI bot traffic — up roughly 300% in 2025 according to Akamai data reported by Search Engine Land — also makes gates 2 and 3 more important than they used to be. With more AI crawlers requesting your pages more aggressively, server response time and robots.txt configuration are gating signals for systems that didn't exist eighteen months ago.
Want a Pipeline Audit on Your Site?
If you want a second pair of eyes on which gate is failing on your highest-priority pages, contact us. We'll run the 10-question audit on three of your pages and send back a one-page diagnostic with the failed gate flagged and the practical next move for each. It's a 30-minute exchange and we don't charge for the initial pass.
If you'd rather start with the foundations, our AEO foundations guide covers the structural and content layers the pipeline assumes are already solid. Most of our clients work through that material first, then run the pipeline diagnostic to identify the specific gates costing them visibility. The pairing — foundations plus a gate-by-gate diagnostic — is what separates AEO programs that compound from programs that grind without traction.
Run the 10-Gate Diagnostic on Your Top Pages
Button Block runs the 10-gate pipeline audit on three of your highest-priority pages as a free first step for Fort Wayne, Auburn, and Northeast Indiana businesses. We flag the failing gate and the practical next move — no rewrite quote required.
Frequently Asked Questions
Frequently Asked Questions
- Do I really need to use the 10-gate AI search pipeline for every page?
- For most pages you don't run through all ten consciously. The 10-gate model is a diagnostic — when a page isn't performing, it tells you where to look. For routine publishing, the practical version is "make sure the infrastructure five pass, then write distinctive content for the competitive five." The framework is a tool you reach for when something isn't working, not a checklist for every post.
- Which gate fails most often for Fort Wayne small business sites?
- In our experience auditing sites in Fort Wayne, Auburn, and across Allen County, gates 4 (render) and 6 (annotation) account for roughly half of the failures we see on small business sites in Northeast Indiana. Both are technical, both are fixable in days rather than months, and both are commonly missed because they don't show up as obvious problems in standard SEO audits. Render failures hide behind reasonable-looking pages; annotation failures hide behind valid-looking schema that mislabels the entity.
- How is this different from a regular technical SEO audit?
- Standard technical SEO audits cover gates 1–5 well. They don't cover gates 6–10 the same way, because those gates are about how AI systems specifically interpret and use the content — annotation accuracy, recruitment into AI candidate sets, grounding in extraction, framing in generation. The pipeline framework extends a technical audit into the AI-specific layers a traditional audit doesn't see.
- Can I run this audit without specialist tools?
- Yes. The 10-question audit uses Google Search Console (free), a curl command (free), Google's Rich Results Test (free), and manual citation checks across ChatGPT, Perplexity, and Google AI Mode (free with consumer accounts). Server log access is helpful for gate 2 but not strictly required for a first-pass diagnostic. The whole audit fits inside an afternoon.
- What's the multiplicative logic and why does it matter?
- In a multiplicative pipeline, the score at each gate effectively multiplies — a near-zero at one gate zeros out perfect performance everywhere else. Barnard's "straight C student" framing captures it: a balanced site that passes every gate moderately outperforms a site with three excellent gates and one failing one. Practically, this means triage upstream gates first; downstream investment doesn't compound until the upstream gate passes.
- Does this framework apply to all AI search systems equally?
- Approximately. ChatGPT, Perplexity, Google AI Overviews, Claude, and Gemini all run through similar sequences — they have to discover content, fetch it, render it, evaluate relevance, and surface a generated answer. The weighting at each gate differs (Perplexity is more citation-heavy, Google AI Overviews leans on traditional ranking signals more, ChatGPT samples broadly), but the gates themselves are consistent. A page that fails at gate 4 fails across all of them.
- How often should I rerun the audit?
- Run it on a focused subset of pages quarterly, and on any new page within a week of publishing. The infrastructure gates rarely change once they pass, but the competitive gates shift as competitors publish, search behavior moves, and AI systems update their weighting. A quarterly review catches drift before it becomes a visibility loss.
Sources & Further Reading
- Search Engine Land: searchengineland.com/10-gate-ai-search-pipeline-find-where-content-fails-476488 — The 10-gate AI search pipeline: How to find where your content is failing (Jason Barnard)
- Search Engine Land: searchengineland.com/why-content-doesnt-appear-in-ai-overviews-473325 — Why content doesn't appear in AI Overviews
- Search Engine Land: searchengineland.com/how-ai-models-understand-your-brand-475993 — How AI models “understand” your brand (Jes Scholz)
- Search Engine Land: searchengineland.com/ai-bot-traffic-surged-publishers-report-473900 — AI bot traffic surged in 2025, publishers report
- Schema.org: schema.org — Schema.org structured data vocabulary
- Google: developers.google.com/search/docs/crawling-indexing/overview — Google Search Central: Crawling and indexing overview