Entity Extractor and Entity Gap Checker
Extract the entities a page names, ranked by salience with the reason for every score. See types with the rule behind them, a section map and co-occurrence, compare against up to three competitor pages for the entity gap, and export about/mentions JSON-LD. Runs in your browser, free.
Compare with competitor pages (entity gap)
Paste a page and press Extract.
What this entity extractor does
An entity is a specific, nameable thing: a company, a person, a product, a place, a defined concept. Paste a page and the extractor finds the entities it names, ranks them by how central each one is to the page, guesses a type for each with the rule that produced the guess, and shows which section each entity sits in and which entities appear together. Paste up to three competitor pages and it runs an entity gap analysis: the entities they name that your page does not, how many of them name each one, and where on their page it appears.
It finishes with an about and mentions JSON-LD block built from the result, with sameAs links you choose. Everything runs in your browser. Several of the entity tools reviewed, including questiondb, wordcount and isocialweb, process text with an AI model on their servers; this one sends nothing.
| Typical entity extractors | This tool | |
|---|---|---|
| Where it runs | Server-side or AI, for questiondb, wordcount and isocialweb | Your browser; nothing is sent |
| Salience | A score, or none | A score with the reasons behind it |
| Types | A label, no explanation | A label with the rule that set it, editable |
| Competitor comparison | None side by side in the tools reviewed | Up to three pages side by side |
| Section map and co-occurrence | Not in the tools reviewed | Yes |
| Structured data | Rarely | about and mentions JSON-LD, sameAs from your links only |
| Limits | Character caps, monthly runs or signup | 500,000 characters, no account |
Tools reviewed in October 2026: WordLift’s demo, isocialweb, questiondb, ralfvanveen, wordcount.com, InLinks and displaCy. Paid suites and APIs may differ.
Entity salience: why position beats frequency
Google Cloud Natural Language describes the salience score as information about the importance or centrality of an entity to the text, from close to 0 for minor entities to close to 1.0 for highly salient ones. Its model is not public, so no outside tool can reproduce it. This extractor uses rules you can read instead, built on the same idea: an entity the page is about shows up in the title, in headings and at the start, not just often.
| Signal | Points |
|---|---|
| Each mention | 1 |
| In a heading, up to two | 5 each |
| In the H1, on top of its 5 as a heading | 4 |
| First named in the opening fifth of the page | 3 |
| Each additional section it appears in | 1 |
The points are divided by the page total, so salience across all entities adds up to 1. A brand mentioned six times in a footer-like closing paragraph scores below a concept named in the title and two headings, which is the right way round. Every row shows the reasons for its score, so you can see when a heading or the opening is doing the work.
How to run an entity gap analysis
- Paste your pageThe HTML is best because it keeps the headings. Markdown or plain text works; lines starting with # are read as headings.
- Paste one to three competitor pagesUse pages that rank for the topic you are writing about, and copy their main text rather than menus and footers.
- Review your entitiesUntick anything that is not really an entity, fix any type that is wrong, and paste a sameAs URL for the ones that have a Wikipedia or Wikidata page.
- Read the entity gapEntities named by two or three competitors, and central on their pages, are the ones a reader expects. Add them where they belong, using the competitor section as a guide.
- ExportDownload the brief for a writer, the CSVs for a spreadsheet, or copy the JSON-LD into your page’s structured data.
Entities, the Knowledge Graph and SEO
Google introduced the Knowledge Graph in 2012 as a way to understand real-world entities and their relationships: “things, not strings”. Naming the specific things your topic involves, and naming them clearly, is how a page tells a search system what it is about beyond its keywords. An entity gap is often the reason a page reads as thin even when its word count is high: the competitors name the tools, standards, people and places the topic involves, and the thin page talks around them.
Structured data can state the same thing explicitly. Schema.org’s mentions property says a creative work contains a reference to a thing without necessarily being about it, and about names its subject matter. Schema App recommends mentions for entities referenced but not explored in depth, and sameAs for linking to authorities such as Wikipedia or Wikidata. That is the split the JSON-LD here uses. Google does not list about or mentions among its rich result features, so treat them as a way to describe the page, not as a ranking switch.
How entities are found and typed
Candidates are capitalised names, including multi-word ones such as Google Search Console or Bank of America, acronyms such as CLS, and a list of known names that are often written in lowercase, such as wordpress or chatgpt. A capitalised word at the start of a sentence counts only if it also appears capitalised mid-sentence, which keeps out words that are capitalised only because they start the sentence. Aliases are merged: Google LLC into Google, and Search Engine Optimization (SEO) into one entity.
Types come from rules, and each row shows which one fired: a known name, a company suffix such as Inc or Labs, a common first name followed by a surname, a place list, or a product word such as App or Console. When no rule matches, the type stays unknown rather than guessed, because a wrong type in structured data is a false statement about your page. You can set any type yourself.
The limits are the limits of rules. It will miss entities written entirely in lowercase that are not on its list, it does not link entities to Wikidata itself, and it only works on English text. For the vocabulary a page shares with its competitors, rather than its named entities, use the Content Gap Checker; to see whether each section of a page can be quoted on its own, use the AI Citability Checker.