AI Search Guide.

For twenty years, a search ended in a list of links. The buyer opened a few, compared, and decided. AI assistants collapse that process: the buyer asks a question and gets a written answer that already names the options, explains the differences and often makes a recommendation.

That changes what visibility means. Ranking fourth on a results page still got you seen. Being left out of an answer means the buyer never hears your name, and they rarely ask a follow-up to find you.

This guide explains how AI answer engines put those answers together, what makes them mention one brand and not another, and how to measure your own position.

Four surfaces account for most AI-assisted buying research today. They work differently, so a brand can be visible on one and absent on another.

Each engine also changes its models and sources often. A snapshot from one day is useful, but trends across repeated, identical checks are what tell you whether you are gaining or losing ground.

Most AI answers about products and services come from two places at once.

• What the model already knows. During training, a model reads a large share of the public web. Brands that appear often, in consistent terms, in trusted places become part of what the model knows about a category.
• What it looks up. For questions about current options, prices or comparisons, most assistants now search the web first, read a handful of pages, and write the answer from them. Those pages are the citations you see.

So there are two ways into an answer: be part of the background knowledge about your category, and be on the pages the assistant retrieves when someone asks. The second responds much faster to work than the first.

No AI company publishes a ranking formula, and the systems change often. What shows up consistently, in the answers we audit and in the platforms' own guidance, is a set of patterns.

• A clear, consistent identity. The assistant needs to be sure what you are. Your name, category, location and offer should read the same on your website, your profiles, directories and press. Conflicting descriptions make an assistant hedge or leave you out.
• Presence where the category is discussed. Assistants lean on comparison articles, review sites, community threads, industry publications and "best of" lists. If those pages name your competitors and not you, the answers will too.
• Pages that answer the question directly. A page that states who the product is for, what it costs and how it compares is easy to quote. A page of slogans is not.
• Comparisons and alternatives. Many buyer questions name a competitor ("alternatives to X", "X vs Y"). Brands with honest comparison content are far more likely to appear in those answers.
• Freshness for current questions. For prices, features and "best in 2026" questions, assistants that search the web prefer recent pages.
• Accessibility to crawlers. If your site blocks the crawlers AI search uses, or hides key information in images and scripts, it cannot be read or cited.

Two questions tell you different things. Ask an assistant "What is [your brand]?" and you learn whether it recognises you and describes you accurately. Ask "What's the best [your category] for [your buyer]?" and you learn whether it recommends you.

Recognition is usually the faster fix, because it depends mostly on the facts available about you. Recommendation depends on how you compare and on where your category is discussed, so it takes longer. A brand that is not recognised is almost never recommended, which is why we fix recognition first.

Typing a few questions into ChatGPT tells you something, but answers vary from one try to the next, and your own chat history can influence them. A useful measurement is systematic.

1. 01Choose the questions your buyers actually ask. Mostly broad category questions, plus comparisons, alternatives, price and feature questions, and trust questions. Avoid questions built around your own unique selling points: they make you look more visible than you are.
2. 02Ask every question on every engine, in a clean session, in the market you sell to.
3. 03Record more than "mentioned or not". Note your position in the answer, whether your site is cited, which competitors appear, and which sources the answer used.
4. 04Check recognition separately by asking each engine what it knows about your brand and comparing that with the truth.
5. 05Keep the questions fixed and repeat the check on a schedule, so changes reflect your visibility and not a different set of questions.

The useful outputs are your share of answers that name you, your average position, how often you are cited, your share of voice against named competitors, and the list of sources that shape answers in your category. That last list is the to-do list.

• Ask each engine "What is [your brand]?" and note anything wrong or missing.
• Make your homepage and About page state plainly what you do, for whom, and where.
• Make sure the same facts appear on your Google Business Profile, LinkedIn page and the main directories in your industry.
• List the ten questions a buyer would ask before choosing you, and check which competitors appear in the answers.
• Find the pages those answers cite, and ask whether you could reasonably be included in them.
• Check that your robots.txt does not block the crawlers AI search relies on.

If you want this done properly across all four engines, with the same questions tracked every month, talk to Recvis. The next guide, GEO Guide, covers how to improve what you find.