Two bars: retrieval, then selection
An AI answer is built in two steps. First the engine retrieves a handful of candidate pages; then it decides which ones to quote. You can be perfectly retrievable and still never get quoted — and you can be highly quotable but never retrieved. Winning means clearing both.
Retrieval is on-page SEO pointed at the questions AI users ask. Selection — will the model choose you from what it retrieved? — is the newer, genuinely GEO-specific problem.
The levers that actually move the answer
| Lever | What to do | Evidence tier |
|---|---|---|
| Title–query match | Make the page title a semantic equivalent of the buyer's question | Cross-niche: cited pages had it, non-cited didn't |
| Inbound internal links | Point links from your other pages into the page you want cited | Our own corpus: ~0% cited at zero links, ~44% at 10+ |
| First-party facts | State specific numbers a model can't generate: prices, speeds, counts | Models quote data they can't produce themselves |
| Freshness | Update with a real substantive change, not a date bump | ~3.2× more citations for genuinely fresh content |
Our internal-link figure is a correlation on our own 91-post corpus at one point in time — strong on our data, not a guarantee for yours.
What we refuse to sell you
Two tactics get sold hard in this category and don't survive contact with our data. FAQPage schema helps a machine parse a page it already retrieved — it does not make an invisible page get cited. An llms.txt file is tidy housekeeping with no measured effect on whether you're recommended.
We ship both because tidy beats untidy. We refuse to bill them as the reason your answer changes, because on our own measurements they aren't.
Common questions
Is there a way to make AI recommend a brand with no reviews or substance?+
No — and anyone promising that is selling you a trick that goes stale. When a rival simply has more substance (more reviews, real credentials, deeper proof), the honest move is to tell you, not to bill you to lose. We only take fights you can win.
Does FAQ schema help me get recommended?+
It's hygiene, not a lever. Structured data helps an engine parse a page it already retrieved; it does not change whether an invisible page gets cited. We add it and don't oversell it.
How long until AI starts recommending me?+
It depends on your starting point and your category, and honest answers here are ranges, not promises. The mechanism is: publish citable pages, get them retrieved and quoted, and re-measure on a frozen panel. Sustained citations correlate with becoming a name models reach for on their own — an edge, not a lock.
keep reading
See what AI says about your brand right now — free, on a live engine.
Scan your domain