Entity Density: The Metric AI Crawlers Care About (and How to Measure It)

Entity Density: The Metric AI Crawlers Care About (and How to Measure It)

Short answer: Entity density in SEO is the ratio of distinct, recognizable entities — people, brands, places, products, and concepts — to the total word count on a page. AI crawlers use it to judge how thoroughly you cover a topic. You measure it by running your text through a named-entity tool like Google's free Natural Language API, then dividing unique entities by words.

Run a page that ChatGPT keeps citing through an entity extractor, then run your own. The gap usually isn't word count or backlinks — it's how many distinct, well-defined things the cited page names per hundred words. That's entity density, and it has quietly become one of the clearest signals separating pages AI engines quote from pages they scroll past. Most content teams still optimize keyword frequency. The engines stopped counting words years ago.

What Is Entity Density in SEO?

An entity is a real-world "thing" a machine can recognize and connect to other things: a person, a company, a place, a product, a date, or a defined concept. Entity density measures how many of those things appear relative to your word count — typically expressed as entity mentions (or distinct entities) divided by total words, times 100.

Google's whole model runs on this. Back in 2012, Google reframed search around "things, not strings" when it launched the Knowledge Graph, shifting from matching letter-strings to understanding the entities behind them. Keyword density counts how often you say "engine." Entity density checks whether you also mention pistons, torque, fuel injection, and spark plugs — proving you actually understand engines, not just the phrase.

For a fuller map of how this connects to AI ranking, our Generative Engine Optimization guide covers the surrounding signals. Entity density is the on-page core of it.

Why Do AI Crawlers Care About Entity Density?

Large language models don't read your page the way a human skims for a keyword. They extract entities and the relationships between them, then decide whether your page is a dense, reliable source on a topic or a thin one padded with adjectives.

That extraction step is measurable. Google's Natural Language API assigns every entity a salience score — a number in the [0,1] range representing how central that entity is to the whole document, per Google's own documentation. A page about "entity density" where the primary entity scores 0.05 is telling the crawler the topic is incidental. One where it scores 0.30+ signals genuine focus.

The correlation with citation is hard to ignore. In one content analysis, pages that LLMs cited averaged roughly 20.6% entity density, versus 5–8% for typical brand copy (MEGA AI) — nearly triple the "things per word." Thin, entity-starved pages don't get quoted because there's nothing concrete to lift.

How Do You Measure Entity Density on a Page? (The Free-Tooling Method)

This is the part almost every "entity SEO" guide skips: the actual arithmetic. Here's a repeatable method using only free tools — no Surfer or Clearscope subscription required.

1. Grab your page's plain text. Copy the body copy (skip nav, footer, boilerplate). Note the exact word count from any word counter.
2. Run it through Google's free Natural Language API demo. Paste the text into the Cloud Natural Language demo — it's free for the interactive tool and returns every detected entity, its type (PERSON, ORG, LOCATION, CONSUMER_GOOD, EVENT), and its salience score.
3. Count distinct entities above a floor. List unique entities with salience ≥ 0.01 to filter out noise. Deduplicate variants ("Google," "Google Inc.") into one.
4. Compute two numbers. Entity density = (total entity mentions ÷ word count) × 100. Entity coverage = distinct entities ÷ (word count ÷ 100). The first tells you concentration; the second tells you variety.
5. Check your primary entity's salience. If your core topic isn't scoring above ~0.10, the page reads as unfocused to the crawler — regardless of density.
6. Benchmark against the SERP. Run the top 3 ranking pages through the same demo and compare. You're not chasing an absolute number — you're beating the pages that already win the query.

A worked example

Take a 1,500-word how-to article. The NLP demo returns 58 distinct entities and 96 total entity mentions. Entity density = (96 ÷ 1,500) × 100 = 6.4%. Entity coverage = 58 ÷ 15 = 3.9 distinct entities per 100 words. Primary-entity salience = 0.22.

Density of 6.4% is on the low side for a tutorial — respectable coverage, but the concentration says you're explaining concepts in long, entity-light sentences. The fix isn't stuffing; it's naming the specific tools, standards, and examples you're already gesturing at vaguely. You can automate this whole comparison inside our AI SEO audit tool, which runs the extraction against your live SERP so you skip the manual paste-and-count.

What's a Good Entity Density by Content Type?

There's no universal target — a product page and a pillar guide play different games. These are the working ranges we anchor to, calibrated against the density analysis above and our own audit set. Treat them as a floor to clear before you benchmark the specific SERP, not a law.

Two rules override the table. First, always beat your top 3 competitors, not the table — the SERP is the real benchmark. Second, coverage beats concentration: a page with 20 different relevant entities outranks one that names the same three entities 20 times. High density with low diversity is just keyword stuffing wearing a new outfit.

How Is Entity Density Different From Keyword Density?

They sound like cousins. They pull in opposite directions.

Keyword density is a metric you can max out and get penalized for. Entity density is one where "more relevant things, clearly named" almost always helps — because it maps to how entity-based SEO actually works. The ceiling isn't a stuffing threshold; it's readability. This is the same shift we broke down in AEO vs GEO vs traditional SEO: the unit of optimization moved from the word to the thing.

How Do You Increase Entity Density Without Keyword Stuffing?

Raising density is about specificity, not repetition. Every time you write a vague noun, you're leaving an entity on the table.

• —Name names. "A popular analytics tool" → "Google Analytics 4." "Recent research" → the actual study and its publisher.
• —Add attributes. Don't just mention an entity — cover its properties, so the crawler sees you understand it, not just reference it.
• —Front-load entities. First-paragraph and H2 mentions carry more salience weight than the same entity buried on screen three.
• —Link entities to related pages. Contextual internal links reinforce the relationships between entities and lift salience across a cluster.
• —Add structured data. Schema explicitly tells engines which entities you're about; the right types materially raise AI citation odds, as we detail in schema markup types AI search engines reward.

The failure mode we see most in audits: teams "add entities" by repeating their brand name and target keyword. That raises count and tanks diversity — the crawler reads redundancy, not authority.

Where Do Most Pages Go Wrong?

The most common problem isn't low density — it's low diversity at high density. A page can hit 14% and still fail because it's the same handful of entities on loop. The second-most-common: strong body copy with a primary entity that never appears in an H2 or the opening, so salience stays low and the crawler misreads the topic.

The third is structural. Entities named in prose but never reinforced with schema, internal links, or consistent naming (five variants of the same product) fragment the signal. AI crawlers build a graph; inconsistent naming hands them broken edges. If you want the machine's-eye view of what's actually landing, our AI Overview Checker shows how engines are currently parsing the page.

Methodology

The framework in this article comes from how we audit growth-stage sites at SEO Magics. For entity analysis, we lean on Google's Cloud Natural Language API for salience scoring, cross-reference detected entities against the top-ranking SERP results for the target query, and validate topical coverage against Semrush and Ahrefs term data. Word-count and density math is computed directly from extracted-entity output, not estimated. The target ranges are working heuristics we've calibrated across client audits and public density analyses — they're starting points to beat a specific SERP, not fixed rules, and we re-benchmark them every time Google or the major LLMs shift how they weight entities. As an AI SEO practice running 12-month optimization cycles, we treat entity density as one input among structured data, internal linking, and crawlability — never a single metric to game in isolation. Where we cite external figures, they link to the original source; where we couldn't verify a number, we've kept the claim qualitative.

FAQ

What is a good entity density percentage?

There's no universal number. Pillar pages tend to work best around 12–18%, tutorials around 10–15%, and product pages 7–12% — but the only benchmark that matters is beating the top 3 pages ranking for your query. Diversity of entities matters more than raw percentage.

Can you have too high an entity density?

Yes, if it comes from repeating the same few entities. High density with low diversity reads as stuffing to a crawler. The safe ceiling is readability — if a human finds it natural and you're naming distinct relevant things, you're fine.

What free tools measure entity density?

Google's Cloud Natural Language API demo is the best free option — it returns detected entities, types, and salience scores. Pair it with any word counter to compute the ratio yourself, and run competitor pages through the same tool to benchmark.

Is entity density a confirmed Google ranking factor?

Google hasn't named "entity density" as a direct ranking factor. What's confirmed is that Google's systems extract and score entities via salience, and that entity-rich, well-structured content correlates strongly with AI citation and visibility. Treat it as a strong proxy signal, not a dial Google published.

How is entity density different from keyword density?

Keyword density counts repetitions of one phrase and can trigger over-optimization penalties. Entity density counts distinct real-world things and rewards breadth. One is a metric to cap; the other is a metric to expand.

Does entity density help with ChatGPT and Perplexity citations?

It helps a lot. LLM-cited pages tend to be markedly more entity-dense than average brand copy, because concrete, named entities give the model something specific to extract and quote. Vague, entity-thin pages rarely get pulled into AI answers.

Ready to see your entity density?

If AI engines are skipping your pages, entity density is one of the first things worth checking — and one of the fastest to fix. Run your top pages through the AI SEO audit tool to see how crawlers are extracting your entities, or if you'd rather have a second opinion on your whole content footprint, book a strategy call. We'll show you exactly which entities the pages that outrank you are naming — and you aren't.