Guide · getting cited in AI answers
Almost every "how to get cited" guide skips the one thing that decides the outcome. Before any tactic, count the distinct domains competing for citations in your niche's AI answers. A fragmented field — dozens of small tools and blogs, no dominant name — is winnable. A field locked up by a few incumbents, Reddit, or big review sites usually is not, and no amount of publishing changes that.
We measured this on two of our own products, run through the identical pipeline. One — a screen recorder — grew to 9% of its niche's AI answers, the single most-cited domain in it, because its niche had ~82 scattered domains with none holding more than 5%. The other — an AI receptionist — sat at 0% for three months, because its niche had 114 domains led by Reddit threads and funded incumbents. Same effort, opposite result, and the deciding number was visible before either product published a word.
Step one isn't a tactic — it's a measurement. Fragmented niche → proceed. Saturated niche → spend the budget somewhere content can't be beaten.
From the pages that did earn citations, three things mattered — and none of them is a trick:
Things people obsess over that our data says don't move the needle:
A common trap: you check Perplexity, see yourself cited, and assume ChatGPT does the same. It often doesn't. In our most recent measurement, our screen recorder was cited by Perplexity across five different buyer queries — and by ChatGPT (GPT-5.5) on none of them except its own brand name. Same pages, same niche; one engine trusts them and the other doesn't yet.
The practical implication: measure ChatGPT specifically — don't infer it from Perplexity. ChatGPT leans harder on established, widely-corroborated signals, so the cluster generally has to be deeper and cited more broadly across the web before it starts pulling you in.
The most common mistake is to ask ChatGPT your question once, see yourself cited, and declare victory. AI answers are non-deterministic. We ran the identical query on the same model twice, back to back — the two answers shared only about a third of their sources, and our own domain appeared in one run and vanished in the next. The live web index shifts and the model samples differently on every call.
So never judge from one run. Run each question at least three times and pool the results, and track the trend month over month. A single reading is a coin flip; the trend is the signal.
Months, not weeks — and only in a winnable niche. Our share climbed over three months of consistent publishing and monthly measurement. In a saturated niche it may never happen, which is exactly why you check winnability before you start.
We saw no evidence they moved our numbers. The levers that actually mattered were niche winnability, cluster depth, and factual accuracy. Don't over-invest in tricks at the expense of those.
No. Engines choose their own sources. Anyone guaranteeing a specific citation for a specific keyword is selling something they don't control.
If your buyers research via AI before they buy, yes — the cited sources are the new page one. If they don't yet, that's a legitimate finding worth knowing, and it saves you the budget.
Measure both, because they diverge. Perplexity tends to cite a broader, fresher set of sources; ChatGPT is more conservative. Winning Perplexity first is common — treat ChatGPT as the harder, later milestone.
We measure your ChatGPT + Perplexity answer-slot share, who holds the slots you're missing, and whether content can actually win it. $300, delivered in 48 hours.
Get your audit — $300