The 10 AI SEO ranking factors that actually matter.
AI engines don't publish their ranking factors. But across thousands of audited queries and four years of testing, the pattern is consistent. These are the ten signals that move AI citations — ranked roughly by leverage.
How to read this list.
These ten factors are ranked by observed leverage — the size of citation movement we see when each is improved in isolation. Your specific niche may differ, but the order is a reasonable starting prior.
Each factor includes the engines it most affects, what "good" looks like, and the practical move that improves it.
Factor 01 — High-authority editorial backlinks.
Engines: All four. Leverage: highest.
Real backlinks from real news editorial — BBC, Daily Mail, FT, Guardian, Forbes, Wired, The Times. Not sponsored content. Not paid placements. Not link-farm output. Real journalists writing about your brand because there's a story.
This single signal moves citations across every engine more than any other. AI engines weight high-authority editorial heavily because (a) they read it during training and (b) they cite it during live web search. PR Backlinks is built to move this exact factor.
Factor 02 — Genuine Reddit footprint.
Engines: ChatGPT (very high), Perplexity (very high), Gemini (medium), Claude (medium-low). Leverage: very high.
Brand mentions in real Reddit threads in relevant subreddits. Not spam, not obvious shilling, not five posts from the same brand-new account. Aged accounts (5–10+ year karma) posting helpful comments in genuine threads, mentioning your brand only when it's the correct answer to the question.
This is the single biggest under-played factor. Most brands ignore it entirely — which is exactly why their AI citations are weak. Reddit SEO works this signal.
Factor 03 — AI-quotable passage structure.
Engines: All four. Leverage: very high.
Content broken into self-contained passages AI can lift cleanly. Each major question gets its own heading, a 40–60 word direct answer immediately beneath, then supporting context. The same format that won featured snippets — now load-bearing for AI citations.
The opposite (long, marketing-flavoured paragraphs that bury the answer in the third sentence) gets cited far less. Restructuring existing content into AI-quotable form is one of the fastest wins for many brands.
Factor 04 — Topical authority across a cluster.
Engines: All four. Leverage: high.
Twenty interlinked pages on a topic beat one landing page by a wide margin in AI citations. Engines reward sites that demonstrate genuine expertise across the topic — cornerstone hub + supporting cluster pages + interlinking between them.
This is why we structure cluster work around hub pages like What is AI SEO? with supporting articles fanning out. The pattern compounds.
Factor 05 — Schema markup completeness.
Engines: Gemini and AI Overviews (very high), Perplexity (high), ChatGPT and Claude (medium). Leverage: high.
FAQ schema, HowTo schema, Article schema with proper author and date markup, Organization schema, Service schema where applicable. Each gives engines structural confidence to lift content cleanly. Pages without it get cited less even when their content is identical.
Treat schema as required, not optional. The implementation cost is low; the leverage is high.
Factor 06 — E-E-A-T signals.
Engines: Gemini and AI Overviews (very high), all others (medium). Leverage: medium-high.
Experience, Expertise, Authoritativeness and Trust. Real authors with bylines and credentials. Dated content. Clear publishing standards. Real organisational identity (not just a brand on a page). Google's longstanding framework, now applied implicitly across every AI engine.
Add author boxes, link them to author pages with bios, mark up author schema, date your content. The improvements compound across all engines because none of them treat anonymous content the same as credentialled content.
Factor 07 — Site speed and technical health.
Engines: Gemini, AI Overviews and Perplexity (high — they crawl in real time). ChatGPT and Claude (medium — they crawl when browsing). Leverage: medium-high.
Sub-2-second load times, server-rendered HTML (not JS-only), valid markup, mobile-responsive, secure (HTTPS). Live-search engines penalise slow or hard-to-parse sites because they have to retrieve and process pages in real time.
If your site is a single-page-app where content only renders via JavaScript, you're invisible to most AI crawlers. Website Development is required for many brands here.
Factor 08 — Content depth and substance.
Engines: All four. Leverage: medium.
Comprehensive, well-researched, original content beats thin AI-spun copy by a wide margin. Engines explicitly weight quality. The signal isn't word count — it's substance: original data, real examples, specific recommendations, expert framing.
Stop publishing 800-word "ultimate guides" written for keyword density. Publish fewer, deeper pages instead.
Factor 09 — Brand mentions (linked and unlinked).
Engines: ChatGPT and Claude (high — they read context, not just links). Perplexity, Gemini, AI Overviews (medium). Leverage: medium.
Even unlinked brand mentions in high-authority publications register. AI engines read the surrounding text, not just the hyperlink graph. A BBC article naming your brand without a link still tells the engines you exist and what context you exist in.
This isn't a metric you optimise directly — it's a side effect of running real PR. But it explains why brands with strong PR programmes outperform brands focused only on link-building.
Factor 10 — Recency and freshness.
Engines: Perplexity (high), Gemini and AI Overviews (medium), ChatGPT and Claude (low — they care about training-data recency, not page recency). Leverage: low-medium.
Pages updated within the last 12–18 months beat older pages on the same topic, all else equal. Especially for queries with a "current information" intent (statistics, trends, recent comparisons).
Build a content refresh cadence. Update dates, expand stale sections, replace outdated stats. The cost is low; the signal compounds across queries.
Two factors that get talked about but barely matter.
- Page-level keyword density. Doesn't move AI citations meaningfully. Engines read context. Stop stuffing.
- "AI-friendly" meta tags. The
llms.txtstandard and similar are not yet weighted by any major engine. Add them if you want, but don't expect citations from them yet.
The practical priority.
Most brands should run factors 1, 2, 3, 4 and 5 in parallel. Those are the highest-leverage moves and the only ones that produce visible citation gains in months rather than years.
Factors 6, 7, 8, 9 and 10 are foundations you should already have. If you don't, fix the broken ones first — then layer the high-leverage factors on top.
That's the structure of every programme we run. See our AI SEO services.
Frequently asked questions.
Are these the same factors across every AI engine?
Mostly yes, with two specific exceptions. The Reddit signal is much stronger for ChatGPT and Perplexity than for Claude. Google's traditional ranking signals (E-E-A-T, schema, technical SEO) carry into Gemini and AI Overviews directly — less so for engines that don't share Google's index. Aside from those, the same ten factors apply to all four engines.
How are these ranked? Where do they come from?
Observed leverage across hundreds of client audits and our own testing. Not a published list from any AI provider — nobody publishes that. The ordering reflects what we see actually move citations when changed in isolation. Your mileage may vary by niche.
Do AI engines have algorithms in the same sense Google does?
Sort of. AI engines have ranking layers that pick which sources to feed into the answer generation step. These ranking layers use signals similar to Google's. The "algorithm" you're optimising for is the combination of those ranking layers plus the model's training data weights — harder to pin down than Google's ranking, but the leverage points are still observable.
Which factor should I fix first?
Whichever your specific gap is — not whichever is "most important" in the abstract. Fast site with no backlinks? Start with PR. Strong authority but no quotable content? Start with AI-Optimised Content. Diagnose your gap before applying tactics in priority order.
How long until ranking factors change again?
The fundamentals (authority, quality, structure, off-site presence) won't change — they're load-bearing for any reasonable AI ranking system. The specific weights and partnerships will keep evolving. Reddit's weight in ChatGPT could drop if the deal changes; new partnerships could raise the importance of new platforms. Build for fundamentals; don't chase tactical edges.