solutions · agencies
Add content retainers without adding writers
Every retainer you win is another writer to brief. IT Master runs a separate pipeline per client site: topics from real search demand, drafts checked by a different vendor's models, articles published to the client's own domain.
reviewed 2026-09-02 · by the IT Master editorial team · how we check facts
- Every retainer we win turns into another writer to recruit, brief and edit, so the margin gets thinner the more we sell.
- Clients ask for eight articles a month and we can staff four before something starts to slip.
- The AI drafts we trialled read like AI, and nobody had the hours to fact-check them before they went live.
- Half our clients run WordPress and the rest run something a developer built, so publishing eats the time we sold as strategy.
- When a client asks what the content actually did, we can show them word counts and not much else.
- Each client site runs without a writer on your payroll.
- Articles publish on the client's own domain, in their own design.
- Drafts that fail the checks never reach a client's site.
- Your team's hours go to strategy, briefs and client calls.
- The monthly report comes from Search Console and assistant citations.
The unit of production is a person, so change the unit
Content retainers scale badly because what you are really selling is a person's week. Each client needs a writer briefed on the niche, an editor to check the writer, and somebody to load the result into a CMS. Win three retainers and you hire. Lose two and you carry the cost until you win them back.
IT Master changes the unit. You pay per published article from a prepaid balance, with no subscription, and a draft that fails the checks is not charged at the full rate. Spend follows what shipped rather than who was on the rota. The customer price is on the pricing page; what you charge your client stays your decision.
Our own generation costs are measured rather than estimated. Median generation cost per article: Lite $0.41, Standard $0.85, Pro $1.46 (our cost, not the customer price).
There are four quality tiers: Lite, Standard, Pro and Premium. A local installer's two posts a month and a competitive B2B client's cornerstone page need not run at the same one.
IT Master charges per published article from a prepaid balance, with no subscription, so an agency's content cost tracks what shipped rather than who was on the payroll that month.
One pipeline per client site
An agency runs client sites, not one site. Each is set up separately: its own delivery key, its own Search Console connection, its own set of topics, and its own body of first-party material.
That separation earns its keep when two of your clients sell the same thing in different towns. They share most of their keyword demand and none of their stock, prices, service areas or support history. Every article is grounded in the first-party data of the client it is written for — product data, datasheets, support questions. That is what stops two client sites converging on the same summary of the same top ten results, and what makes either one hard for a competitor to copy.
Topics come from real demand on that client's own property: keyword data plus their Search Console, rather than a content calendar carried over from another account.
Sites we run on the engine ourselves include cctvline.co.uk and cctvmaster.co.uk (UK CCTV, trade and retail), eurocctv.co.uk (UK CCTV trade) and 1xai.ir (AI developer tools, in Persian).
On IT Master each client site is set up on its own: its own delivery key, its own Search Console connection, its own topics and its own first-party data.
Published to the client's own site, in their own design
Articles publish on the client's domain and render in the client's own templates. There is no network of ours that the content sits on instead.
Each article arrives finished: JSON-LD schema, internal links to the site's other pages, a hero image and an FAQ block. IndexNow then pings the search engines rather than waiting on a crawl.
Delivery follows whatever the client already runs:
- WordPress plugin: install it, paste one key, and articles arrive as posts with the Yoast or Rank Math fields filled in.
- @itmaster/sdk for Next.js: articles render inside the site's own components, so a developer-built site needs no CMS bolted on.
- REST pull feed for any stack that would rather fetch than be pushed to.
- Signed push webhook for a headless CMS or a build pipeline, signed so the receiving end can verify it.
Shopify and Webflow we set up on request. Wix and Squarespace clients get a copy-paste queue instead, where a finished article is pasted rather than written.
The checks protect your reputation before the client's
An agency pilot with generated content usually stalls over one paragraph: a confident, wrong claim on a client's site that the client finds first. The answer here is structural rather than a promise. Writers are Claude models. The fact-check and novelty judges are Gemini and GPT, so the model that writes is never the model that checks.
Four checks run on every draft: a fact-check, a novelty score against the competitor pages fetched for that topic, an E-E-A-T critique, and AI-tell detection. A failing draft goes back to the writer with the specific objections, for up to three revision rounds.
The pass rates are published rather than implied. Across 186 Standard-tier runs, 51% of drafts passed every check first time; the rest were revised or rejected. Pro-tier judges are stricter: 25% first-pass, across a small sample of 16 runs.
A rejection is the system working: what fails does not publish. How it works sets out the full sequence.
In IT Master's validation gauntlet the model that writes is never the model that checks: writers are Claude models, and the fact-check and novelty judges are Gemini and GPT.
The seat you keep is editor-in-chief
Taking the writer out does not take the accountable human out. It moves them to the seat that needs judgement. One named person owns what a site publishes, sets the standards and answers for the output. At IT Master that person is Rodd Azad, a software engineer and the founder. At your agency it should be the strategist who already fronts the account.
That seat works in hours a month rather than hours a day:
- set the standards and the wording each client expects;
- steer the topic clusters each client works through;
- read the first few weeks closely, then move to sampling;
- judge which topics are worth the spend and which are not.
Note where the gate sits. The validation checks are the gate: an article that passes publishes, one that fails does not. Human judgement goes into standards, topic direction and reporting rather than into a queue of drafts waiting to be waved through, because a queue nobody clears is how content programmes quietly stop. The editor-in-chief entry sets out the role in full.
What goes in the monthly report, and how to pitch it
Reporting is where content retainers are lost. The engine measures what a client recognises, from that client's own Search Console: impressions, clicks, position, and which pages moved. A nightly loop watches for articles that are decaying and queues them for a refresh rather than leaving them to rot.
It also measures the part most agencies cannot yet answer. The same nightly loop asks GPT and Perplexity the questions each article was written for and records whether the client's page is cited, so the citation rate is a number in the report instead of a claim in a deck. Nobody can promise a citation from an assistant. Showing whether it happened is the honest version.
For the pitch itself, the free tools run on a prospect before you quote: the site check, the AI visibility checker, the AEO checker, and the AI crawler check, which shows whether the assistants' crawlers can reach their site at all. Start with one client, measure a quarter, then decide how many follow.
questions people ask
Do the published articles look like ours or like yours?
They publish on your client's own domain and render in that site's own design, so the page belongs to the client. Nothing in the article points a reader to an intermediary, and there is no network of ours that the content sits on instead. The account you administer, where client sites are set up and reporting is read, is an IT Master account. Most agencies treat that the way they treat any other tool in the stack: you own the strategy, the reporting and the client relationship, and the engine is production capacity behind it.
How are two clients who sell the same thing kept apart?
Each client site is set up separately, with its own delivery key, its own Search Console connection, its own topic set and its own body of first-party material. Every article for a site is grounded in that site's own product data, datasheets and support answers, which is where the real differences between two similar businesses live: stock, prices, service areas and the questions their customers actually ask. Two competitors in the same trade therefore end up with genuinely different pages rather than two rewrites of the same top ten results.
What happens when a draft fails the checks, and do we pay for it?
The draft goes back to the writer with the specific objections from whichever check failed, for up to three revision rounds. If it still fails it does not publish, so nothing that failed reaches the client's site. You pay per published article from a prepaid balance, and a draft that fails the checks is not charged at the full rate. For a sense of how often this happens: across 186 Standard-tier runs, 51% of drafts passed every check first time; the rest were revised or rejected.
Which platforms can our clients be on?
WordPress clients install the plugin and paste one key, and posts arrive with the Yoast or Rank Math fields filled. Next.js sites use the @itmaster/sdk package, so articles render in the site's own components. Any other stack can pull from the REST feed or receive a signed push webhook, which covers headless CMS setups and static builds. Shopify and Webflow we set up on request. Wix and Squarespace clients get a copy-paste queue, where a finished article, schema and all, is pasted rather than written.
What approves an article before it goes live?
The validation gauntlet is the gate. A draft that passes the fact-check, the novelty score, the E-E-A-T critique and AI-tell detection publishes; one that fails does not publish at all. That is the trade an agency is making. Instead of a queue of drafts waiting for a human to read them, every draft is checked by models from a different vendor than the writer, and your hours go into standards, topic direction and reporting. Agencies usually read closely for the first few weeks on a new client, then move to sampling.
Will the articles read as written by a machine?
AI-tell detection is one of the four checks, run by a model from a different vendor than the writer, alongside the fact-check, the novelty score against the competitor pages fetched for that topic, and the E-E-A-T critique. Failing any one of them sends the draft back with specific objections. The other half of the answer is grounding: articles are written from the client's own product data, datasheets and support questions, so they carry specifics that a model writing from general knowledge could not produce.
related
See what search engines and AI assistants find on your site
Free, no account. Type your address and we show you what is missing and what we would write first.