Reference

What Chompr checks.

Every item in a Chompr result, explained in plain language: what it looks at, why it matters and what to do about it. Click any name in the extension to land on its entry here.

The score

The big number in the middle of the result. It is built from one pillar, Factual, and then reduced if AI search crawlers are blocked.

Score (0 to 100)

Scored

Shown as: Well-optimized for AI retrieval · Needs work · Critical issues detected

What it is: A readiness score for the page you are looking at. Green is 70 and up, amber is 40 to 69 and red is below 40. The same page always gets the same score, and every result records the algorithm version that produced it.
Why it matters: It gives you one number to track while you fix a page. It is not a prediction of whether an AI engine will cite the page, and a higher number has not been shown to mean more citations. Read it as a checklist with a number on it.
How to improve it: Work on Factual first, since it is the only pillar that moves the number. If the result says crawlers are blocked, fix that before anything else.

Factual

Scored · 25 pts

What it is: How concrete the page is. Pages with specific figures, lists and links to sources a reader can check score higher than pages made of general statements, even when the general statements are better written.
Why it matters: In a controlled experiment published at KDD 2024 (the GEO study), adding statistics, quotations and source citations to web content made it noticeably more visible in AI-generated answers. Factual is built on those findings, which is why it is the pillar behind the score.
How to improve it: Replace vague claims with real numbers, dates and measurements. Put steps, options and comparisons in lists or tables. Link to the original source for the facts you state, such as official documentation, standards bodies or published research.

AI crawler access

Whether AI search engines are allowed to fetch the page at all. This is the one part of Chompr that has been tested against real citation data.

AI search crawlers

Affects the score

Shown as: Blocked from OpenAI · Perplexity: allowed (one row per engine)

What it is: One row per AI search engine (OpenAI, Anthropic, Perplexity and DuckDuckGo) showing whether your site's robots.txt lets its search crawler in. Each blocked engine takes points off the score.
Why it matters: In our study of AI search citations, pages were cited about half as often by an engine whose crawler they blocked. How much a block costs differs between engines, so the penalty does too. Where we had no data for an engine, the result labels its penalty as an untested default.
How to improve it: If you want to appear in that engine's answers, allow its search crawler in robots.txt. The result names the exact crawler to allow.

Not reachable by AI search

Affects the score

Shown as: CAPPED · Score capped

What it is: The page tells search engines not to index it or not to show snippets from it, using a noindex or nosnippet robots directive. When that happens the score is capped at 39 no matter how good the content is.
Why it matters: A page that asks not to be indexed or quoted is asking to be left out of AI answers too. Nothing else on the page can make up for that, so the cap makes the problem impossible to miss.
How to improve it: Remove the directive from the page's meta robots tag or X-Robots-Tag header. If this is a staging or preview page that is meant to be hidden, turn on pre-publish mode instead.

Crawler policy

Informational

Shown as: Opted out of AI training · User-initiated fetches disallowed

What it is: Your robots.txt stance on AI crawlers that are not search crawlers: the ones that collect training data, and the ones that fetch a page when a user asks an assistant to read it.
Why it matters: Blocking these is a legitimate choice and does not affect the score. It is shown so you can confirm your policy says what you meant it to say.

The checklist

Five areas of page hygiene, out of 75 points in total. They come with fixes but do not change the score. The evidence behind them is weaker than for Factual, so we list them as things worth doing rather than count them.

Schema

Checklist · 15 pts

What it is: Whether the page describes itself in structured data (schema.org JSON-LD), and whether that description is complete enough to be useful: what the page is, who published it and the details that type of page needs.
Why it matters: Structured data tells a machine what a page is without having to guess from the text. Microsoft has said Bing and Copilot use schema.org markup to understand page content.
How to improve it: Add one JSON-LD block that matches the page, such as Article, Product, Organization or SoftwareApplication, and fill in the properties that type expects. One complete entity beats many half-filled ones.

Structure

Checklist · 12 pts

What it is: How cleanly the page is organised: one clear main heading, sections that nest properly and paragraphs of a readable size.
Why it matters: Retrieval systems split pages into passages along headings and paragraphs. A clean outline keeps each passage about one thing. No published study has yet tied structure to AI citations, which is why it is on the checklist rather than in the score.
How to improve it: Use a single H1 for the page topic, H2s for the main sections and H3s inside them. Use semantic elements such as <main> and <article>, and break up walls of text.

Entities

Checklist · 8 pts

What it is: Whether the page names things or only gestures at them. "Postgres 16" is specific, while "this tool" is not.
Why it matters: A passage that names its subject is easier for a retrieval system to match to a question about that subject. A passage full of "it" and "this" only makes sense next to the text around it.
How to improve it: Use full proper names for products, people, companies, places and versions, especially at the start of a section. Repeat the name where a reader landing on that paragraph alone would need it.

Answerability

Checklist · 20 pts

Shown as: Answerable

What it is: How easy it is to lift a clear answer from the page: a short, self-contained answer near the top, headings phrased the way people ask questions and a length that suits the topic.
Why it matters: In one large study of ChatGPT citations, nearly half of the cited passages came from the top 30% of the page. AI answers quote passages that make sense on their own, so an answer buried halfway down, or one that depends on earlier paragraphs, is easy to miss.
How to improve it: Open with a brief answer to the question the page exists to answer, written so it makes sense when quoted alone. Phrase some H2s as the questions your readers actually ask, and answer each one in the first sentences beneath it.

Retrieval

Checklist · 20 pts

What it is: The basics every search system reads first: a descriptive title, a meta description, a canonical URL, a declared language and a summary worth quoting.
Why it matters: Search engines document the title and meta description as inputs to the snippets they show. A canonical URL tells crawlers which copy of the page is the real one, and a declared language helps route the page to the right readers.
How to improve it: Write a title that says what the page is about, not only the site name. Add a meta description that summarises the page in a sentence or two, a rel="canonical" link and a lang attribute on the <html> element.

Other things you may see

Notes and tools that appear in the result when they apply.

Quick win

Informational

What it is: The single fix Chompr thinks is worth doing first on this page, with the points it would add. Checklist fixes are marked as checklist points so they are not mistaken for score points.
Why it matters: A list of every possible fix is hard to act on. The quick win puts the most valuable one at the top.

Low confidence

Informational

Shown as: LOW CONFIDENCE · This may not be the actual page

What it is: Chompr suspects it scored something other than your real content, such as a bot check, a cookie or challenge screen, a login redirect or a page with no readable text.
Why it matters: If the page you see is not the page a crawler gets, the result describes the wrong page. The warning does not change the score. It tells you not to trust it yet.
How to improve it: Get past the screen and rescan, or check whether crawlers are being shown that screen too. If they are, that is the problem to fix.

Pre-publish mode

Informational

Shown as: DRAFT · Production origin · Draft vs live · Rendered vs raw

What it is: A mode for checking drafts on staging, preview or localhost. It scores the page as it will be once it is live: the staging noindex no longer caps the score and becomes a ship-blocker item instead. You enter your production origin so Chompr checks the robots.txt that will actually apply.
Why it matters: Staging sites are almost always set to noindex, so in normal mode every draft would show as capped. Pre-publish mode separates "this is a draft" from "this page has a problem". You turn it on yourself; Chompr never switches it automatically.
How to improve it: Clear the ship blockers before you publish. Two optional comparisons are available: Draft vs live compares the draft with the page that is live now, and Rendered vs raw shows whether your main content is in the HTML or only appears after JavaScript runs.