Answer Radar / Content gaps

Content gaps: what your page is missing that the AI answers had

A recommendation gap tells you an engine chose someone else. The content gap view tells you which of your pages was supposed to answer, and names what it does not say that ChatGPT, Gemini, Perplexity and Claude agreed the answer needed.

Start tracking Gap analysis that survives noise

Updated 2026-09-04

What the content gap view does

It reports what is missing and writes no advice about it. That part is yours.

For each measured question, the engines' admissible answers are reduced to a requirement set: phrases at least two different engines used, plus the products their judged mentions named. Each page in your project's evidence store is scored on which requirements it contains, whole word, with links stripped. A question no page covers above the ceiling is flagged as a content gap and the closest page is named, because extending a page that is almost there usually beats writing a new one.

The real response, run on 4 September 2026 on our own project

We are printing the refusal rather than a mock. Our linkeddit.com project had no pages in its evidence store when we ran it, and the tool said so.

answer_content_gaps_get({
  project_id: "<your project id>",
  run_id: "19910a90-4d5d-4549-8b7f-1116184f9b5a",
  only_gaps: true,
  mode: "summary"
})

The exact call through the Linkeddit MCP connector. project_id is your project; the run id is ours.

Response, verbatim fields

measured: false
reason: no_tracked_pages
page_count: 0
source: answer_source_documents + answer_samples + answer_mentions
thresholds: min_providers 2, min_requirements 5, gap_ceiling 0.5, max_requirements 40, mention_parser_version mentions-v4
upstream: answer_gaps_list on the same project returned nine open recommendation gaps. Rank 1: the prompt about competitor monitoring software for pricing changes and product launches, intent 0.85. Rank 2, Crayon alternatives: Linkeddit neither mentioned nor cited in four successful samples.

A view that scored zero pages and returned coverage 0.0 would have looked like data. This one returned a reason. The same rule applies to a page whose text cannot be read.

How a page is scored

Four rules, each one a decision we made so the number means the same thing next month and for the next customer.

A requirement needs two engines

A phrase counts only when at least two different providers used it in their answers. One engine sampled five times repeats its own phrasing; two engines agreeing is the better question. Products come from the judged mentions: the alternatives the answers actually named.

Three things are excluded

The prompt's own words, because every answer to a question contains the question. Publications and review sites, because they are cited sources, not products. Your own brand, because requiring your page to mention you is free coverage for the pages that need work most.

Matching is whole-word, URLs stripped

The page's extracted text has its links removed, then each requirement is matched as a whole word. Coverage is binary per requirement: the page says it or it does not. A term that appears only in a link slug is not covered, because a slug is not prose an engine reads as content.

Refusals, not zeros

A page whose text could not be read is unusable with the reason, never coverage zero. A prompt whose requirement set came out empty is unusable for the same reason. A rate with an empty denominator is absent from the response rather than 0.0.

The thresholds, as the tool reports them

Every response carries its own disclosure block. These are the values on 4 September 2026.

Field	Value and meaning
min_providers	2. Engines that must have used a phrase before it is a requirement.
min_requirements	5. Below this the prompt's requirement set is unusable, not scored.
max_requirements	40. The cap per prompt, so a long answer set does not drown a page.
gap_ceiling	0.5. A prompt is a content gap when no tracked page reaches this coverage; the closest page is named.
tracked page rule	The newest fetched document per URL belonging to the project. A page must be in the evidence store before it can be scored.
mention parser	mentions-v4. Quoted with every coverage number so two numbers are only compared when the instrument matched.

Start with only_gaps for the questions no page covers. Then the low coverage band for the pages that are almost there. Then draft a fix against the gap, which pulls the same evidence, and let verification say whether it moved.

What a coverage number is not

Copied from the notes the tool returns with every response.

A missing requirement is an observation about text. It is not advice, and it does not assert that the page ought to mention the thing. Deciding that, and writing it, is yours.
Coverage is not a ranking claim and not a prediction of being cited. It says what the page contains relative to what the answers contained.
Requirements are pooled across engines deliberately: what answers to a question say is a claim about the world, not about one instrument. Each requirement reports how many engines used it.
A coverage number is reproducible only when quoted with its document id and its run or window. Re-fetching the page changes what was scored.

Frequently asked questions

What is a content gap in AI search?+

A content gap is a buyer question that none of your pages covers to the level the answer engines' own answers did. Answer Radar derives a requirement set from the admissible answers to each tracked question, phrases at least two engines used plus the products they named, scores each of your stored pages against it, and flags a question as a gap when no page reaches the coverage ceiling. It names the closest page so you extend rather than start over.

How is this different from a recommendation gap?+

A recommendation gap says an engine recommended someone else, or nobody, for a question. On our own project on 4 September 2026 the gap list held nine of them, ranked by intent, and the second read: Linkeddit was neither mentioned nor cited in four successful answer samples for the prompt about Crayon alternatives. The content gap view asks the next question: which of our pages was supposed to answer that, and what does it not say that the answers did.

Why did the view return no result on your own project?+

Because we had not put any of our pages into the project's evidence store when we ran it. The tool returned measured false with the reason no_tracked_pages, and we have printed that response rather than a mock. Pages enter the store when an audit fetches a cited source or when you fetch one deliberately with answer_sources_fetch.

Does coverage predict whether I will be cited?+

No. Coverage is an observation about text relative to what the answers contained. It is not a ranking claim and not a prediction. Why an engine chose a source is on the methodology page's list of things no measurement here can tell you.

Does it tell me what to write?+

It tells you what is missing, by name, and stops. A missing requirement is an observation, not advice that your page ought to say the thing. Whether to name a competitor on your page is your editorial call, or your assistant's, which can read the same evidence through the connector.

Can I filter it to one page or one question?+

Yes. Filter by url to see one page against every question, by prompt_id or prompt_text to see every page against one question, by only_gaps for the questions no page covers, or by a min and max coverage band for the pages that are close. Scope to a run or a date window. Evidence mode returns the per-requirement verdicts.

Find out which of your pages was supposed to answer, and what it left out

The content gap view is part of Answer Radar, included with the Compete plan, in the browser and through the MCP connector.

See pricing How Answer Radar works