You know which page should be cited. Your pricing page, your comparison page, the service page that sells. ChatGPT cites a competitor instead. A page fix tells you why, on that exact page, and what to change to make it citable.
Why does a good page never get cited by AI engines?
Because an AI engine does not read a page the way your visitors do.
Your visitors see a clean layout, prices, a clear offer. An AI robot may see a 403 because your firewall turns it away. It may get an empty page because the prices only appear after JavaScript runs. It may read your offer but find nothing that says where you operate, for whom, or how much it costs, and move on to a page that does.
None of this shows up in your analytics. The page simply never appears in the answers, and nobody on the team knows where to start. A page fix is the place to start.
What do you see when you open a page fix?
Open any page of your site from Site health and run the fix. About a minute later, the page shows three things.
What an AI understands from the page. One line: who, what, for whom, where. An empty box is already a finding: "Where: not found on the page" means an engine cannot recommend you for a local question.
A to-do list sorted by severity. Blocking first, then AI impact, consistency and SEO hygiene, the same levels as Site health. Each task opens on why it matters, the facts found on the page, and what to do.
The tools to do it on the spot. JSON-LD blocks ready to copy, the pages to link from with the anchor to use, what disappears without JavaScript, a suggested title or meta description.
A switch moves from the guided view to the expert view: the raw data for whoever implements it, with a JSON export to hand to a developer.
What does the page fix check?
It starts from the problems the monthly crawl already found on the page, then reads the page live, three times: as a browser, as OAI-SearchBot, ChatGPT's search robot, and fully rendered.
Access. Is the page indexable, which AI robots does robots.txt allow, does the page answer the AI robot like it answers a browser, is there a bot wall or an X-Robots-Tag in the way.
Rendering. What the page says without JavaScript compared with what it says once rendered: headings, prices and structured data that only exist after the script runs.
Structured data. The JSON-LD type the page should carry for what it is (a product, an article, a local business), the required fields missing, values that contradict the visible text. Mentionable then writes the blocks to add: the expected type, a breadcrumb, and an FAQ when the page has questions and answers. Values it cannot know are marked « … » for you to fill.
Links and duplicates. The internal links that point to the page, the closest pages that do not link to it yet, and other pages of your site competing for the same prompts.
Freshness. Whether the page says when it was last updated, and whether it has gone stale.
What the engines look for. Mentionable compares the page with your brand profile and with the searches the AI engines actually ran on the prompts tied to the page. Missing facts quote your own profile word for word (service area, prices, audience, opening hours). Missing topics quote at least one real search from an engine. Nothing is guessed.
Does Mentionable rewrite the page?
No, and that is deliberate. Your offer pages stay in your words.
Mentionable tells you which information is missing and which sections to plan. When the title or the meta description is flagged, it suggests one, drawn from the page alone and within the right length, to read before publishing.
When your Action plan schedules a full rework of a page that sells, the page also carries the rework brief: the angle, the target outline, the queries to cover, the links to add and the pages to merge into it. Your team writes; Mentionable says what the rewrite has to achieve.
How do you know the fix was applied?
You do not have to report it. At the next crawl, Mentionable runs the fix again on every page whose fix is not fully applied yet, and compares it with the previous one.
A missing point is fixed. Missing facts and topics are judged by reading the new version of the page. A suggested internal link is checked on the page it was meant to come from. The result reads "4 items fixed out of 6".
Fully applied, the page is checked off in your Action plan on its own and written into the project's action journal. Partly applied, the journal records what was done, and the page stays open with what is left.
Did the fix change anything in AI answers?
This is the question your director will ask, and the page answers it with numbers.
Once a page is fixed, a "Fix impact" block appears at the top of it. Mentionable takes the prompts tied to the page and compares two rates, four weeks before the fix against four weeks after: how often AI answers mention your brand, and how often they cite this exact page.
The change is set against your other prompts over the same dates. From one month to the next, a large share of the domains AI engines cite changes on its own: without that baseline, market noise would pass for the effect of your fix. Mentionable waits for enough scans on each side before showing a figure.
The table then splits the result by engine, with each engine's usual delay next to it: a few days for Perplexity, one to three weeks for ChatGPT, four to eight weeks for Google AI Overview. If Google Search Console is connected, the page's Google impressions and clicks are compared over the same periods.
Who can run a page fix, and what does it cost?
Page fixes are included in every plan, with no monthly limit, on any page of the latest crawl. A page gets one fix per crawl, and the next crawl checks it again.
Anyone on your team with edit rights can run one. On an agency workspace, a client with a read-only account sees the result but cannot run it.
Your own AI can do it too. Through the Mentionable MCP server, the run_page_fix tool launches a fix and get_page_fix reads it back, verification and impact included.
Page fixes are one part of Site health: the crawl finds the problems across the site, the fix solves them page by page.