Make the documentation you already wrote findable

The problem is almost never that the answer is missing. It is that it is on page four of a search result nobody scrolled. Ask in plain language, get the passage, with the link.

Crawl your docs free Read the crawl guide

No credit card. Nothing to migrate.

They asked

“How do I add another login?”

No words in common

Seatsmatched passage

Ask in their words: 'An extra login' finds the page you titled 'Seats'.
The passage, not the page: The relevant part, quoted, with a link to the full article.
Nothing to migrate: It crawls what you already publish, wherever it lives.
Gaps, ranked: The questions your docs never answered, ordered by demand.

A printed article on a desk in hard window light, one paragraph run through with yellow highlighter and another boxed in pen with an arrow to a handwritten margin note.

Good documentation, unusable search

Keyword search only works if the reader already knows your vocabulary. Customers do not call it a 'seat', they call it 'an extra login'. The article exists, matches nothing, and a support ticket gets filed about a documented feature.

Tickets arrive about things the docs explain in full
Search returns the changelog for almost every query
Readers use their words for your features and find nothing
Nobody knows which articles are missing, only that something is

What changes when documentation becomes answerable

Finding an answer usually means already knowing your vocabulary, your information architecture, and which of forty articles is the right one.

Meaning, not keywords

Matches on what a question means, so a paraphrase reaches the right passage even with no words in common.

Answers, not ten blue links

The relevant passage is quoted back, so nobody skims a three-thousand-word article for one line.

Always a citation

Every reply links its source, which sends readers into your documentation rather than away from it.

Current without a ritual

Scheduled re-crawls re-index only what changed, so a published edit reaches the assistant on its own.

The gap report

Every question your docs failed to answer is logged; on Growth and above they are grouped and ranked into a writing backlog.

Any language

Answers arrive in the language of the question, from content you maintain in one.

Ask in your words, not in ours

Matter Chat matches on meaning rather than on exact strings, so a question phrased in a customer's vocabulary still reaches the passage written in yours. The reply is the answer itself, with a link to the article it came from.

Semantic matching: 'How do I add another login' finds the page about seats.
The passage, not the page: The reader gets the paragraph that answers them, not a link to go hunting in.
Always a citation: Every answer links its source, so the reader can keep going in the real doc.
Crawls what you already publish: Point it at your help centre and docs site. There is nothing to migrate.

A widget conversation in which a visitor asks how much it would cost to send a bag to France, and the assistant answers with the tracked rate, the free-delivery threshold and a numbered citation. — The question says “send a bag to France”. The article it came from is titled “Shipping”. Keyword search finds nothing here; this does.

The gap report is the point

Every question your documentation does not answer is recorded, on every plan. On Growth and above they are grouped by demand as well, so instead of guessing what to write next you get a ranked list of what readers actually asked for and did not get.

Ranked by real demand: The most-asked missing answer sits at the top of the list.
Phrased as they asked it: You get the real wording, which is usually a better article title than the one you would have chosen.

A content-gap report built from failed answers, listing topics such as product origin, subscription management and international shipping, each with the visitor's own wording and a link to write an answer. — Built from answers that failed, kept in the reader's wording, and refreshed weekly.

Stays current without a maintenance ritual

Documentation rots quietly. Matter Chat re-crawls on a schedule, so an edit you publish on Tuesday is answerable without anyone remembering to press a button.

A source list showing four indexed pages, each marked ready with its chunk count and how long ago it was updated, above a refresh-all control. — Every source carries when it was last read. The refresh is there for when you cannot wait for the schedule.

The honest part

A search box that fails beats an assistant that guesses

Documentation is the one place where a confident wrong answer does real damage: the reader believes it, acts on it, and files a ticket about something else entirely. So the retrieval either finds a passage in your content or the assistant says it does not have one.

Answers come from your pages, not from the model's memory
No matching passage means it says so, and offers a human
Every claim carries the link a reader can check
The refusal is recorded, and becomes the next thing to write

Which means coverage is the ceiling on quality. A thin help centre produces a bot that refuses often — correctly, and unhelpfully. The gap report is how that gets fixed.

A woman writing in a notebook at a tall desk against a large industrial window, morning light flooding past her.

Passage found

The answer, in your words, with the link.

Shipping ↗

Nothing found

“I don't have that in the documentation.” Then a human, if you offer one.

No source, no answer

Making an existing help centre answerable

No migration, no re-authoring, and no separate system to keep in sync. You are pointing a reader at content you already wrote.

The add-source row offering five ways to add knowledge — crawl a site, a single URL, an uploaded file, Notion, or a written snippet — with a domain field ready for a crawl. — A help centre goes in as a crawl. Anything not published on the web goes in as a file or a snippet.

1
Add your docs as a source
Point the crawler at your help centre and your marketing site. Anything not published on the web can be uploaded as a document instead.
2
Prune the noise
Tag archives, paginated listings and author pages dilute retrieval. Read the source list once and remove what is not an answer.
3
Test the vocabulary gap
Ask questions the way a customer would, not the way your docs are titled. That gap is exactly what this is meant to close.
4
Work the gap report
Write the top item, wait for the next crawl, and watch that question stop arriving. That loop is the whole operating model.

Before you point it at your docs

Short answers. If yours is not here, the assistant on this page will try it — and tell you honestly if it cannot.

Do I have to move my documentation anywhere?

No. Matter Chat crawls your published site — help centre, docs, marketing pages — wherever they live. Nothing needs to be exported, re-hosted, or rewritten in another format.

What about content behind a login?

The crawler reaches what is publicly published. Material behind authentication can be added as uploaded documents instead, so internal or gated content is still answerable.

How is this different from the search my docs platform ships with?

Built-in search matches words. This matches meaning and returns an answer rather than a list of ten links, then cites the article so the reader can verify it. The gap report is the part no search box gives you at all.

All solutions