What CiteGraph does

Who it is for

Founders, marketers and SEO consultants who want a straight answer to “does AI recommend us?” and a plan when the answer is no. Solo founders use it on one domain. Agencies track several client domains on one account and send each client a link to its report. It is not for people who want a rank to buy: positions are measured from real answers and cannot be edited by anyone, including us.

What a tracked buyer question is

A question a real buyer types into an assistant, in their words, not a keyword. “Best invoicing tool for a freelancer who hates spreadsheets.” “Rewardful vs FirstPromoter for a bootstrapped SaaS.” “Cheapest way to get DPDP compliant without a developer.” CiteGraph reads your site and writes ten of these for your product and category. You see them before anything runs, delete the ones that miss, and add your own. The set is then fixed, so week-to-week movement is real movement and not a different question.

How it finds the pages engines cite, and the answers you are missing from

Each question goes to the engines on your plan (all four on Starter and Growth) through their official APIs with web search on, the same retrieval a buyer triggers. We keep every answer verbatim. Two things are counted separately: named, your product appears in the answer’s text, and cited, your site is one of the pages the engine used as a source. A product can be recommended constantly while its own words never reach the buyer, so both matter.

Every cited page is recorded and graded by what it actually says about your category. Across every scan we have run, the pages engines cite are mostly not the brands’ own sites: Reddit threads, listicles, review platforms and directories. Those pages are your gatekeepers, and the map of them is the most useful thing in the product. Every question where you are not named is a lost answer, listed with the verbatim answer and the reason the engine picked someone else.

What a fix includes

For every lost answer, the one sentence that would have put you in the answer, in your product’s words. An action plan in order: each move names the target page, why it matters, and the first step, and the next move follows you around the app until it is done. Site fixes: fifteen checks on whether the engines can reach, read and quote your site, from robots.txt for the eight crawlers that feed them to schema and the plain sentence that says what you are, each with the code ready to paste. And pages: for a question you lose, the page the engines would cite, written from the verbatim answers, the pages that won and your own site, every sentence traced to its source and a marked gap where your site is silent. Plus the shorter drafts: the pitch to a listicle owner, the answer to a forum thread.

Which assistants are covered

ChatGPT, Claude, Gemini and Perplexity. For ChatGPT we run the OpenAI model with live web search through the API rather than the consumer app, which is why an answer here can differ from one typed into the chat window. Claude and Gemini are the production models through their official APIs; Perplexity returns its citations natively. We never scrape a chat interface and never simulate an answer. Starter and Growth ask all four; ChatGPT Only asks ChatGPT alone.

Methodology and how often it runs

On Starter and Growth: ten questions, four engines, three runs each, 120 sampled answers per scan. A scan you start on Growth runs five: 200. ChatGPT Only asks its 20 questions on ChatGPT alone, three runs each. Rates carry Wilson 95% intervals, and a change is only reported as movement when it exceeds the combined margins of the two scans. Scans re-run weekly on every plan. Category leaderboards are dated snapshots (the latest from August 2026) and cannot be bought or edited. The full method, including what we deliberately do not do, is on the methodology page.

What it costs

A site readiness check is free and needs no signup at citegraph.app/readiness. Weekly tracking starts at $19 a month for 1 project on ChatGPT alone (ChatGPT Only), then $29 for 2 projects and $69 for 3, both asking all four assistants. Every plan is billed monthly and carries a 120-day guarantee. The full table is on the pricing page.

Questions