How the engine works — Generative Engine Optimization (GEO)
The Ranking Factory asks 4 of 6 AI engines per run — your pick of ChatGPT, Gemini, Perplexity, Claude, Grok and DeepSeek — the questions your buyers ask, finds where they name someone else, writes the missing evidence from your own material, publishes it as a page on your own site, and asks again to see whether it is cited.
GEO = Generative Engine Optimization — being the source AI answer engines name and cite. Every citation you are shown is read out of a real AI answer, with the number it was counted out of.
Every project runs the same loop. On Growth and above it runs on its own every few hours; on Starter it runs when you press Run.
Each run asks the engines you choose the questions of your market — some with your name in them, some without — and records, per engine, who is named, cited and recommended. Every rate carries the number it was counted out of. Google AI Overviews are read as a separate surface, on the searches you choose to watch.
Where the engines name someone else, or answer without evidence you already have, it opens a gap: the subject, what is missing, and what would close it. Each cycle it finds new topics in recent answers and retires the ones measured again and again without ever naming you.
Before a word is written, two or three of your own AI providers agree what the answer has to be — a comparison, a guide, a checklist, a calculator — and which elements it must contain. The shape comes from the question, not from a content calendar.
The writer works from your own pages, videos and statements, and cites only outside sources it has fetched and read. A product of yours that nothing of yours explains is not written about: you are asked what it is first.
One page per subject on your own site — WordPress, Shopify, Wix, Ghost or any webhook — assembled from its parts and, where your platform allows it, updated in place when new material arrives. Held for your review if you ask for that, and always on a page an AI engine is citing. Once your site serves its own IndexNow key file, new pages are also announced to Bing and the indexes that read it.
The next run asks again. Each gap shows how many of the runs since publishing cited what was published for it, and with your Cloudflare zone connected you see which AI crawlers fetched the page. The loop keeps running.
A citation lands on a page. So the work ends on pages AI engines retrieve — your own site first — with the supporting pieces around it, each carrying a different kind of evidence: the answer, the data, the proof.
One page per subject, answer first, every section able to stand on its own when it is quoted, sources cited, and schema.org markup that records when it was drafted and published and what it drew on.
A canonical About page on your own site: what the business is, what it offers, who runs it and where each fact can be checked. Its markup lists your profiles as your identities and your published pages as pages about you.
The explanation each page is assembled from, with its data and deck where they are made, published so that every one can be opened and checked on its own.
Switch it on for a campaign and explainer videos go to your own YouTube channel, a source Google’s AI Overviews draw on heavily.
Update posts to your Google Business Profile, for businesses people visit.
Your own facts stated in another place, each carrying the same record of when it was drafted and who published it. Still your word, not anyone else’s.
Consistency, not copies.
AI search fans one question out into many and answers from what it can retrieve, so the same article posted five times adds nothing. The page, the data behind it, the video and the profile all agreeing about the same business is what makes you citable — and whether you are cited is read out of the next AI answer, never assumed.
The engine that does the publishing — SmartStack — also keeps it working, and tells you what it did.
Every campaign action is scored before it runs. High-confidence work goes first, and no single property is leaned on too hard.
Every night it checks what was published — Drive permissions, Blogger limits, cloud pages that stopped answering, expired tokens — and rebuilds what broke.
Five gauges per project: evidence coverage, gaps closed out of gaps found, the score from your latest visibility run, how often that run’s answers cited you, and the change since the run before. A gauge with nothing behind it shows a dash and says what is missing.
Per-keyword positions with target ranks, and the search features on each page of results — featured snippet, People also ask, and the map pack where DataForSEO read the page — steering what gets built next.
Every piece built, repaired and checked, in order. Reviewable at any time.
One email per site every Monday, on by default, with every campaign on the site in it: what was cited, what was published and done, what needs you. One click turns it off.
One email per site every Monday, on by default. Switch a site off in Settings → Weekly report, and every email has its own unsubscribe link. An account that had the old emails is sent a notice the Monday before it starts, naming every address that will get it.
How a citation is counted: how we measure.
Every plan runs the same loop. Growth and above run it on their own; Growth comes with a 14-day free trial.
One client, on demand
14-day free trial
Multi-client
Enterprise-grade
Find where the engines name someone else, publish the page that answers the question on your own site, and see in the next run whether it is cited.