Be easy to find when customers ask AI assistants.

Under the hood

The agent is designed to work in rounds. Each is a scheduled job with its own id and a fixed list of tasks: refresh the questions, check every published answer against the current fact sheet, draft what is new or out of date, and run the monthly sample. Every sentence on a published page is linked to the fact it rests on, so when a fact changes, the pages that use it are found at once and marked for a redraft.

Under the hood

Nobody can see what a person types into an AI assistant, so the agent works from questions it can read: the search queries Google Search Console reports for your site, your Google Business Profile reviews, the enquiries you share, and last month's sample of what assistants say. The language model groups questions that mean the same thing and matches each group to an existing page, so a weak page is improved rather than copied. You see the map and can strike out anything you do not want to be found for.

Under the hood

The fact sheet is a list of confirmed facts: services, areas, the prices you choose to publish, and qualifications and memberships you can show, each with who confirmed it and when. Every claim in a draft is linked to a fact; a claim without one is cut and becomes a question for you. Structured data (labelled facts, in a shared standard format, that search engines read) is generated from the same facts, so the visible page and the markup cannot disagree.

Under the hood

The preview shows the page, its structured data in plain words, and every claim with its source fact beside it. Publishing is a separate step that only runs with your approval of that exact version: the approval is tied to a fingerprint of the content, so if a word changes afterwards the page waits for you again. Your corrections go back into the fact sheet, so a fact is fixed once, not page by page.

Under the hood

The page is built as plain HTML that reads in full without scripts, so search engines and assistants that fetch pages can read every word. The structured data is checked against the format's rules before it goes out, internal links are added from related service pages, and the updated sitemap is submitted through Google Search Console. Every publish is recorded with the version you approved, so a page can be rolled back to the one before.

Under the hood

The question set is agreed with you and kept the same from month to month, so results compare like with like. Each answer is stored word for word with the date, and the language model reads it to record whether your business was named, how it was described and which sources were cited. A wrong description, such as an old price or an area you no longer cover, becomes a task for the next round.

Under the hood

The report puts the sample beside the Google Search Console data for each answer page: how often it appeared in search, and for which questions. It quotes the assistants rather than summarising them, so you can read exactly what was said about you. Nothing in it is a promise that an assistant will mention you. It shows whether the work is being picked up, and pages that are not go to the top of the next round.