AI search does not share one Google-style ranking algorithm. Engines mention you, cite a URL, recommend you, or omit you — not the same as Search Console position. On 8 Oct 2026 this URL was HTTP 200 with an April layout, 0 Search Console page rows in 28 days, and Crawled — not indexed (last crawl 2 Jul 2026). Keep Search Console next to the checker.
First published 6 Apr 2026. Full rewrite. April copy is labeled history only — do not reprint its unauditable multi-query study framing.
What we tested (8 Oct 2026)
Unique data is this URL’s Search Console row, URL Inspection, live HTML, and today’s Google SERP — not a roundup’s paraphrase of engine internals.
- This URL, live before rewrite: HTTP 200, canonical
https://scope.online/learn/ai-search-ranking-factors. Layout date still 6 Apr 2026. Body ~184,067 bytes of April shell. No datedWhat we tested (DATE)box. April framing claimed a multi-thousand-query “ranking factor” study without a primary sample you can audit on this page. - URL Inspection, 8 Oct 2026: coverage Crawled — currently not indexed. Last crawl 2 Jul 2026 (
2026-07-02T11:42:11Z). User canonical = Google canonical = this URL. Referring URLs included the glossary sibling and the sitemap. Indexed ≠ present; crawled ≠ ranking. - Search Console, this URL, LAST_28_DAYS: no rows (0 impressions attributed to the page in the pull). Page-level absence is not proof the H1 is worthless — it is proof Google has not yet treated this URL as a ranking document after the July crawl.
- Search Console, site totals (prior morning freeze, 7 Oct 2026): 28 clicks, 4,676 impressions, CTR ~0.6%, average position ~28.4. Site totals are truth; page breakdowns under-count.
- Google SERP, desktop US, 8 Oct 2026:
ai search ranking factors— Scope absent. Page one was SEO/GEO operator guides (SEOcrawl AIO factors, Google Search ranking systems doc, WebFX, Finch, Rankability ChatGPT factors, Wellows AIO factors). B2B marketer intent — owner-SERP pass.ai ranking factors 2026— Scope absent. Mix of AIO factor guides, a 155-factor listicle, and one Stanford “AI Index” collision (wrong job if you only chase that title).llm seo ranking factors— Scope absent. LinkedIn posts and “LLM SEO” listicles.what makes chatgpt recommend a business— Scope absent. Reddit, agency blogs, OpenAI community threads.
- Answer engines this run: chatgpt.com 403, perplexity.ai 403. gemini.google.com and grok.com returned HTTP 200 HTML shells without a citation list we could trust. Record blocked / shell, never invent citations.
- Sitemap: live
/sitemap.xmlstill lists this<loc>among 341 URLs. Sitemap inclusion did not flip coverage after the 2 Jul crawl.
The gap is not “missing the secret 200-factor checklist.” It is an April shell that Google crawled and left unindexed, while page one for the H1 is already owned by operator guides that separate Google AI features from assistant recommendations.
What “ranking factors” means when there is no ten-blue-link SERP
Google Search still has ranking systems. Google documents those systems for classic results and is explicit that AI features in Search still require the page to be indexed and snippet-eligible — not a special machine-readable file. That is one scoreboard.
ChatGPT, Claude, Gemini, and Perplexity are different products. They retrieve, compress, and synthesize. A useful operator model is four stages, not one weighted formula:
- Gate — Can a retrieval bot fetch the money URL? Is it indexed where the engine depends on a public web? Does
robots.txtallow the bots you care about? - Extract — Does the first HTML state name, category, who you serve, and one proof in plain text? Or is the answer trapped behind client JS?
- Entity — Is the same business described the same way on-site and off-site, or does the model confuse you with a homonym?
- Consensus — Do independent pages repeat the same fact, or is your site the only place the claim appears?
When people say “AI ranking factors,” they usually mix (a) what Google AI Overviews / AI Mode cite from the open web with (b) which brands assistants name in a shopping or vendor answer. Measure both. Do not pretend they share one coefficient table.
Related reading: GEO vs SEO, how to get cited in AI search, and how Claude recommends businesses.
How does this differ from classic Google SEO?
Classic SEO optimizes for queries, crawl, index, and rank on Google’s results. That work still matters. AI features in Google Search lean on pages that already qualify for Search. If the money URL is not indexed, you are not in the pool.
Assistant answers add a second failure mode: omit. You can rank on Google for a related query and still never be named when a buyer asks ChatGPT who to hire. You can also be mentioned without a citation, or cited on a URL that is not the page you wanted to monetize.
Practical split:
| Signal | Google Search / AIO | Assistants (ChatGPT, Claude, …) |
|---|---|---|
| Primary log | Search Console | Prompt panel: mention / cite / recommend / omit |
| Index requirement | Hard gate for organic + many AI features | Still need a public page when browsing is used |
| On-page extract | Titles, headings, snippet-eligible copy | First 40–80 words and quotable facts |
| Off-site | Links, brand queries, entity consistency | Lists, docs, UGC mix differs by engine |
| Success metric | Clicks, impressions, position | Named in the answer the buyer sees |
Use both. Scope’s product measures the assistant panel; Search Console remains the Google panel. See GEO vs SEO for the full framing.
Which factors actually show up in operator tests?
Treat the list below as a checklist of failure modes, not a ranked algorithm leak. Every item maps to something you can inspect without inventing engine weights.
Gate factors
- Index coverage on the money URL. URL Inspection: submitted and indexed, self-canonical, not soft-404. This URL failed that gate on 8 Oct (crawled, not indexed).
- robots and bot access. Allow the retrieval bots you care about. Blocking the open web while hoping for citations is a self-own.
- Stable HTTPS URL. No accidental noindex on the template that ships the answer.
- Server response the bot actually gets. A 200 HTML document with the answer in the first response beats a spinner shell.
Extract factors
- Direct answer under the H1 in 40–80 words. Models and snippet systems both prefer a complete sentence early.
- Question-form H2s that match how buyers ask, not only how marketers title slides.
- One proof fact (dated test, count, comparison row) that stands alone if quoted out of context.
- SSR or prerender so the entity sentence is in the first HTML, not only after hydration.
- Visible updated date when the claim is time-sensitive — layout date and body date should not disagree after a rewrite.
Entity factors
- Same name + category everywhere. Homepage, product page, LinkedIn, directories.
- Disambiguation for homonyms. “Scope” collides with education services, health brands, and project-scoping agencies on Google. Entity sentences reduce wrong-job retrieval.
- Money page match. If engines cite
/aboutwhile you sell from/pricing, fix the page they can extract — or earn the citation on the URL you want.
Consensus factors
- Independent pages that repeat the fact. Listicles, docs, partner pages, credible reviews.
- Engine-specific sources. ChatGPT leans more on UGC and lists in many categories; Claude’s retrieval mix is not identical. Do not optimize only for Reddit if your buyer asks Claude.
- Fresh corroboration. A 2024 roundup that omits you still shapes answers until something newer repeats a better set.
For a methodology-shaped scoreboard rather than a factor myth, see the Scope AI visibility index notes.
Worked example: this URL on 8 Oct 2026
Prompt job: an operator googles ai search ranking factors looking for what to change on-site and off-site so AI answers name their product.
What Google showed: operator guides and platform pages. Scope’s own guide URL was absent from those SERPs the morning of the rewrite.
What Search Console showed for this URL: no 28-day page row.
What Inspection showed: crawled 2 Jul 2026, still not indexed.
Operator read: shipping a long “factor encyclopedia” in April did not create an indexed ranking document. Before arguing about weights, clear the gate. After the gate, the page must carry a dated test and a checklist a marketer can run in thirty minutes — not an unauditable study claim.
What we will re-check after deploy: HTTP 200 with needle What we tested (8 Oct 2026), layout date 8 Oct 2026, sitemap still containing this loc, and a later Inspection crawl. Indexation can lag days to weeks. Do not extra-submit the sitemap the same morning as a Learn rewrite unless last download is already stale on a separate technical gate.
Counter-example: the 155-factor checklist without a prompt panel
A common failure mode on the SERP is the giant unordered list: hundreds of “AI ranking factors,” many copied from classic SEO checklists, with no measurement loop.
Why it fails operators:
- No omit/cite/recommend log. You cannot tell whether a change moved ChatGPT, Claude, or only your feelings about completeness.
- Collapses Google and assistants. A Core Update-style item and a “get listed on a blog roundup” item are not the same lever.
- Invites fake precision. Publishing “factor #47 weight 3.2%” without a primary sample is the same class of error as inventing industry-wide win rates.
- Ignores gate failures. Polishing meta descriptions on a crawled-not-indexed URL wastes the week.
If a vendor sells a factor score without showing the prompts, engines, and dates behind it, treat it as content marketing — useful as a brainstorm list, useless as a scoreboard.
What should you ignore or downgrade?
- Special AI-only schema as a ranking cheat. Google has been clear that AI features in Search do not require extra AI text files or novel schema beyond normal eligibility. Schema can help extractability; it does not mint citations by itself.
llms.txtas a ChatGPT rank hack. Treat it as optional documentation for humans and some crawlers, not a guaranteed mention lever.- One-engine anecdotes. A single ChatGPT answer on one laptop is not a factor model.
- Homonym traffic. If impressions are for another company’s name, rewriting your H1 will not help AI recommendations of you.
- Net-new thin URLs while money URLs stay CNI. Fix or upgrade the URL that already matches the H1.
30-minute operator checklist
Run this on one money prompt and one money URL.
- Write the buyer prompt exactly as a customer would ask (tool category + use case). Save it.
- Search Console → URL Inspection on the URL you want cited. Note coverage, last crawl, canonical.
- Fetch first HTML (
curlor View Source). Confirm name, category, audience, proof appear without executing JS. - Check robots for the bots you care about. Note allows/disallows.
- Run the prompt on ChatGPT, Claude, Gemini, Perplexity (or record blocked). For each: omit / mention / cite / recommend. Save cited URLs.
- Diff cited URLs vs your money URL. If they cite a competitor list, open it. If they cite your
/blogfrom 2023, decide whether to upgrade that page or earn a better source. - One on-page fix only: tighten the first 80 words or add a dated proof table — not a full redesign.
- One off-page action only: update or pitch a real list/doc that already ranks for the category — not a spray of directory spam.
- Re-ask the same prompts in 7–14 days. Compare the log. No log = no learning.
- Keep Google SEO running. Indexation, titles, and internal links still feed the pool AI features draw from.
When you want a structured scan across prompts, use the AI visibility checker. For agent-readable monitoring, see Scope MCP for AI agents and /agents.
FAQ
Q: Is there a single list of AI search ranking factors?
A: No reliable public list assigns stable weights across Google AI features and every assistant. Use stage checklists (gate, extract, entity, consensus) and measure omit/cite/recommend on your prompts.
Q: Do Google ranking factors still matter for AI Overviews?
A: Yes for eligibility. Pages generally need to be indexed and snippet-eligible. That is necessary, not sufficient, for being named in an overview or an assistant answer.
Q: Will more words rank me in ChatGPT?
A: Length without extractable facts and off-site consensus rarely moves mentions. A clear entity sentence plus corroborating sources beats a 5,000-word brochure that never states who you are.
Q: How is this different from SEO ranking factors?
A: SEO factors target Google’s results. AI answer work adds retrieval and synthesis failure modes. Keep Search Console. Add a prompt panel. Do not replace one with the other.
Q: What should agencies track weekly?
A: For each priority prompt: engine, date, mention kind, cited URL, and whether the cited page names the client. Pair with Search Console clicks on the money URL.
Q: Can I skip indexation if I only care about ChatGPT?
A: Risky. Browsing-capable products still need public pages. Unindexed money URLs also fail Google AI features. Clear the gate first.
Labeled history (April 2026 shell)
The 6 Apr 2026 version framed “AI search ranking factors” as a large multi-query study and long factor narrative. That framing is retired. This rewrite keeps the slug because it already matches the B2B H1 job and sits in the sitemap. It does not restate unauditable percentages or study sizes from the April body.
Next steps
- Run the 30-minute checklist on your top three revenue prompts.
- Upgrade the money URL engines already touch before minting new slugs.
- Track Google and assistants as parallel scoreboards — checker, citation tracking, compare Scope next to your SEO suite.
AI search “ranking factors” are less a secret formula and more a disciplined way to stop failing gate, extract, entity, and consensus — while Google SEO keeps the pages eligible in the first place.