Learning how to rank in AI search starts with an uncomfortable fact: most engines do not rank pages in a visible list, they cite a handful of sources inside an answer. These 15 playbooks show what you can control for each engine, in the order that is least likely to waste your time.
What "ranking" means in an AI answer
In classic search you can point to position four and say you hold it. In an AI answer you are either one of the linked sources or you are not, and the set of sources may change when the question is phrased slightly differently. Some engines show a numbered list of citations. Some show a few cards. Some mention a brand without linking at all.
That changes how you measure. Instead of a rank, you track three things: whether your page appears as a cited source, whether your brand is named in the answer text, and whether anyone clicks through. We explain how to log these by hand in every playbook, because few engines give you a report.
The order of work
Every playbook in this section follows the same priority logic, adapted to the engine. We put access first because it is binary, and measurement last because it needs everything else in place.
| Order | Question to answer | Typical action |
|---|---|---|
| 1 | Can the engine's crawler or fetcher read the page? | Review robots.txt, CDN bot rules, server-rendered HTML |
| 2 | Does the page answer a real question in a quotable passage? | Rewrite the opening, add a direct answer near the top |
| 3 | Does it contain details only you can supply? | Add first-hand data, named sources, dates, prices |
| 4 | Is it fresh and honest? | Update facts, show the real date, remove stale claims |
| 5 | Can the engine understand who you are? | Author pages, organisation details, consistent naming |
| 6 | Can you see what is happening? | Manual prompt log, referral analytics, webmaster tools |
Where an engine documents something that changes this order, the playbook says so. For instance, an engine with a documented crawler gives you a clear on and off switch. An engine that reuses Google's index means your normal Google SEO work comes first.
What we deliberately leave out
You will not find fixed percentage weights for "citation factors" in these guides. Some vendors publish them. We have not found one that is confirmed by the engine itself, and optimising for a number nobody can verify is a good way to waste a quarter. When a playbook mentions a figure, it names the source and tells you how much weight to give it.
You will also not find tricks. Hidden text, fake reviews, mass-produced thin pages and invented statistics may fool a weak system for a week. They also violate the guidelines of every engine we cover, and they make your pages worse for the human who finally lands on them.
Choosing where to start
Not every engine deserves equal effort. A rough way to decide:
- If your traffic comes from Google today, begin with Google AI Overviews and Google AI Mode. They reward work you may already be doing.
- If your customers ask assistants for product or service recommendations, read ChatGPT and Perplexity.
- If you sell in China, or to Chinese-speaking users, read DeepSeek, Qwen and Kimi.
- If you care about privacy-conscious audiences, Brave Search is worth an afternoon.
Before you commit, read the methodology so you know how much confidence each guide deserves. Then work through the playbook for your first engine, log your results for a month, and only then add a second one.