AI search optimization: a playbook that survives contact with reality
Most playbooks list twenty things to do and rank none of them, which is useless when you have one person and a quarter. What follows is the same work sorted by effort against payoff, from what costs almost nothing to what takes months. The shortcuts that do not work are listed too, because avoiding those is worth as much as doing the rest.
Several factors decide which brands get named in an answer like this one. Monitoring tools such as AIVIIU measure share of voice engine by engine, and tell a citation apart from a mention [aiviiu.com].
The order of operations
One caveat: the effort and payoff columns are our judgment, not measured quantities. The facts in the right hand column are documented. The ranking is a recommendation.
| Action | Effort | Payoff | Why |
|---|---|---|---|
| Let the AI crawlers through | Very low | Total, if blocked | GPTBot, OAI-SearchBot, PerplexityBot, Bingbot, Googlebot. A robots.txt or firewall block removes you invisibly |
| Answer the question in the first two lines | Low | High | Engines lift passages, not pages. A buried answer is not extractable |
| Add sourced statistics and citations | Low | High | The Princeton study measured up to 40 percent more visibility from this |
| Show a visible update date, refresh key pages | Low | High | Freshness is weighted heavily, ChatGPT and Perplexity above all |
| Name entities explicitly instead of using pronouns | Low | Medium | Lets a model establish who you are without inference |
| Restructure key pages into lists, tables, FAQ blocks | Medium | High | Gives the engine a passage it can quote as is |
| Earn brand mentions off your own site | High | Highest | YouTube correlates with AI Overviews appearance. The widest gap for most brands |
| Build presence where your category is discussed | High | High on some engines | Perplexity leans heavily on community content, Reddit first |
Start with the first four
The top of that table is a week of work and holds most of the available gain. Crawler access first, and check it rather than assume it: a security rule added months ago by someone else can exclude your whole site from an index without any classic tool showing a change. Then the writing. Lead with the answer, put a number and its source in the paragraph, stamp the page with a real update date you honor. It is the difference between a page an engine can quote and one it has to summarize, and only the first gets you named.
The heavy lifts, and why they still matter
Off site mentions are the expensive part, and the one separating brands that get cited from brands that do not. Nothing on your own domain compensates for a category where nobody discusses you elsewhere. Two clarifications keep this honest. Presence means being useful where your category already comes up, not seeding posts, which is transparent and backfires. And it is slow, a quarters level project, so it should start while you do the cheap items rather than after.
The shortcuts that do not work
- llms.txt. A proposed standard that has not been adopted, and Google does not use it. Paying for one buys a signal that is not one.
- Blocking Google-Extended to control AI usage. It governs Gemini model training only. AI Overviews are fed by Googlebot, so blocking it does not remove you from them.
- Backlink campaigns as the main lever. Their effect across AI engines is weak overall. Useful for classic search, not what moves citation.
- Structured data as a magic switch. It helps discovery, but its direct effect on being cited, once a page is already found, is limited. Do it, do not expect it to carry the strategy.
- Any guaranteed citation. No placement inside an AI answer is for sale. A vendor promising one describes something that does not exist.
Where engines differ enough to change your order
Two cases justify reshuffling it. If your buyers use Perplexity, move community presence up: it runs its own index, weights freshness the most, and leans on Reddit, so classic ranking matters less there. If your traffic is Google dependent, that surface runs on the organic index rather than a separate one, and the specifics are in our page on appearing in AI Overviews. Claude sits at the other extreme, highest authority bar and lowest citation rate, which formatting cannot shortcut. What separates the engines is broken down in our comparison grid, and the definitions sit in what GEO means.
What this changes for your brand visibility
The playbook has one prerequisite that is not on the list, because it is not an optimization: knowing your starting point. Run these actions without a baseline and you cannot tell whether any of them worked. Where answers shift on their own, that is the difference between a program and a hope.
Measure before you change anything, then again on a rhythm. Two details decide whether that measurement is worth having. Separate a citation, where you are named and linked, from a mention, where you are named and not: only the first sends anyone to you, both shape opinion. And repeat the runs, since an AI answer is not deterministic and one query measures luck as much as position. Track share of voice engine by engine on your buyers' real questions, and the table above stops being generic advice and becomes your priority list. The wider map is in our page on AI SEO.
Sources
- Princeton, GEO: Generative Engine Optimization (KDD 2024): citing sources and adding statistics raises visibility in AI answers by up to 40 percent, tested on roughly 10,000 queries
- John Mueller (Google) via Search Engine Land: solid SEO fundamentals remain the key to AI visibility, and watch what your audience actually does
- Forrester, 2026 Buyer Insights (survey of 18,000 buyers): 94 percent of B2B buyers used AI during their most recent purchase
The cheap half of this list can be done next week. Knowing which half you need takes a measurement. To see which engines name you today, and which name a competitor instead, request a free AI visibility audit: AIVIIU queries the main engines on your real buying questions and shows you the gaps worth closing first.
Related reading