Cover your category's questions in the form assistants quote

Every release note describes a change somebody needed, and behind that need there is usually a search. Most notes stay three lines long and never become the article that would have caught it.

IT Master grounds each draft in your own first-party material, wherever that material exists:

• product and release information — what changed, what it replaced, from which version;
• your documentation — real parameter names, limits and error codes;
• support answers, and the questions behind them, from your helpdesk or your community.

From that base it writes pages a competitor cannot copy: a migration guide naming the removed endpoint and its replacement, an explainer quoting your real rate limits instead of a round number, a troubleshooting page assembled from the ten support questions one feature generates. Specifics of that kind are what the novelty score rewards, because they appear on none of the competitor pages fetched during research.

Where your material is thin, the article says less rather than filling the gap from a model's memory. First-party data covers how the grounding works and what happens to a claim nothing can source.

When a developer asks an assistant which tools handle a job, the answer is assembled from pages the assistant retrieved seconds earlier. A page earns a place in that answer when it states the answer in a self-contained passage, under a heading phrased the way the question was asked, on a site the assistant already associates with the category. That is answer engine optimisation, and it is a structural property of the page rather than a trick.

Every article is built for that retrieval step: the direct answer in the opening lines of each section, a FAQ block carrying FAQ schema, Article JSON-LD, and headings phrased the way people ask the question. IndexNow pings the search engines as soon as a page is live.

Then it measures. The nightly loop puts the questions each article was written to answer to GPT and Perplexity, and records whether your page is cited, so the citation rate is a number you watch rather than a hope. Nobody can promise a citation. We publish in the shape assistants quote and report honestly on whether they do.

One piece of that surface is worth naming because developers ask about it. The llms.txt convention — Jeremy Howard's proposal, “adding a /llms.txt markdown file to websites to provide LLM-friendly content”, structured as an H1, a summary blockquote and H2 sections of annotated links (llmstxt.org, checked 4 September 2026) — is rendered for you from your published articles, so the file is an index that stays current rather than a stub somebody wrote once. Whether any given assistant reads it is not settled, and we do not claim it is.

Measure your current site first with the free AI visibility checker, AEO checker and AI crawler check.

Most SaaS marketing sites run Next.js or something custom, and most content tools assume WordPress. IT Master ships through whichever door your site already has.

• @itmaster/sdk for Next.js: install the package, add a route, and published articles render inside your own components, typography and navigation. Articles arrive as data, so nothing is pasted into a CMS by hand.
• REST pull feed: any stack fetches published articles with a per-site key, at build time or per request.
• Signed push webhook: the engine posts each article to an endpoint you control, signed so you can verify it before you store it. This is the route for a headless CMS, a static generator or your own pipeline.
• WordPress plugin if the blog itself runs WordPress: install it, paste one key, and posts arrive with Yoast or Rank Math fields filled.

Shopify and Webflow we set up on request; Wix and Squarespace get a copy-paste queue. Whichever door it comes through, an article carries its JSON-LD, internal links, hero image and FAQ block, so the site receives a finished page rather than a body of text to dress.

The failure that matters to a technical audience is a confident paragraph that is wrong: a parameter that does not exist, a limit lifted from a competitor's docs and attributed to you. The answer is structural rather than a promise. Writers are Claude models, and the model that writes is never one of the judges.

Four checks then run in parallel. None is run by the model that wrote the draft, and every check includes at least one model from a different vendor: Gemini and GPT lead the fact-check and the novelty score, while the E-E-A-T and AI-tell panels pair a Claude judge with a GPT or Gemini one. The four are a fact-check against the retrieved sources, a novelty score against every competitor page fetched during research, an E-E-A-T critique, and AI-tell detection. Fail one and the draft goes back to the writer with the specific objections, for up to three revision rounds. A draft that still fails is rejected, and is not charged.

Across 186 Standard-tier runs, 21% of drafts cleared every check on the first attempt with no revision, and 51% got there within the three revisions allowed; the rest never published. Pro-tier judges are stricter: 25% cleared every check across a small sample of 16 runs.

The four tiers — Lite, Standard, Pro and Premium — differ in the writer model and in how strict the judges are. Fixed prices per published article: Lite £9, Standard £19, Pro £39 and Premium £69 — about $11, $24, $50 and $88, charged in GBP. Length does not change the price, only the tier does; the tiers are compared on pricing. For pages carrying API-level detail, Pro is the tier we suggest, so budget £39 an article rather than £19 for those. The whole flow is on how it works.

No SaaS or developer-tools tenant is live yet, so rather than a case study from your own category, here is what the same pipeline has done on a UK CCTV site it runs today. Treat it as evidence of the mechanism, not a like-for-like result.

CCTV Master, a UK security-camera retailer, took impressions from 143,000 to 471,000 over 28 days across 36 published guides, with 1,011 of its 1,196 URLs indexed.

Developer tools is a harder category for this: it is crowded and heavily documented, official docs and long-established blogs usually hold the top of the results, and a rewrite of the existing top ten has nothing to add. That is exactly what the novelty check is built to catch — it sends the draft back to your own limits, releases and support threads for something the other ten do not have.

The developer tools page goes deeper into that category, and the free site check shows which of your category's questions your site currently leaves unanswered.

Five situations where we would tell you to spend the money elsewhere.

• Your category has no search demand yet. If you are creating a category, the queries that would carry this do not exist. Discovery reads DataForSEO and your own Search Console, and it will report a thin list rather than invent one. Design partners and founder-led sales are the play at that stage.
• Your product changes weekly and your docs lag it. Grounding is only as good as the material it retrieves. If the documentation is behind the product, articles written from it will be behind it too, and the refresh loop reacts to Search Console decay rather than to your release notes.
• You need legal or security review of every published word. Review-every-article is a setting, so this is possible — but it is a person in the loop on every piece, which is the cost the rest of the product exists to remove. Weigh that before buying it.
• You want the docs themselves written. This is the layer above your reference, not the reference. Your docs stay the source of truth and the articles link into them.
• You are buying it to be cited by ChatGPT by a specific date. We publish in the shape assistants quote and we measure the citation rate nightly against real queries. Nobody can promise the citation, and we will not.

There is also a practical constraint on languages. Which ones we can take on depends on the search data available for that market, so ask before you start rather than after the first batch.