How to Improve Your Brand's Visibility in AI Search: Seven Moves, in Order
Seven moves that get a brand named and cited in ChatGPT, Perplexity and Google's AI answers, in the order we would make them, and what does not work.
By Norbert Szabo, founder of Vercatus
Published · 7 min read
To improve your brand’s visibility in AI search, make sure AI crawlers can read your site, find the pages the answers cite for your topics and earn a place on them, answer the shortlist questions plainly on your own site, keep your facts consistent everywhere, and judge every change before you scale it. The order matters: the first moves remove blockers, the middle ones do most of the work, and the last one tells you which of them worked.
In short
- Measure first: your share of AI answers per topic and per engine, next to the brands named instead.
- Fix access before anything else. Of 64 well-known websites we probed in September 2026, 27 served their home page to none of the three AI assistant crawlers we sent.
- Most of the effect is on other people’s pages. Brands named on many of the pages engines cite are named far more often in the answers.
- Then answer the comparison, alternative and pricing questions on your own site, in passages that can be quoted alone.
- Change one thing at a time and compare it with topics you left alone. Answers drift on their own, so before-and-after charts flatter everything.
Before you start: know where you stand
You cannot improve what you have not measured. Take a baseline of your share of AI answers on the questions your buyers ask: category questions (“best payroll software for a small agency”), comparisons, alternatives, pricing and integrations. Record it per engine, because ChatGPT, Perplexity, Gemini and Google’s AI Overviews and AI Mode often name different brands, and keep the brands named instead and the sites cited.
Vercatus’s free audit does this once on four engines; the same page shows how to run a manual check in an afternoon. What makes the numbers trustworthy explains why many different questions beat a few repeated ones.
1. Make sure AI crawlers can read your site
An engine cannot cite a page it cannot fetch. When we probed 64 well-known websites on 18 September 2026, 27 served their home page to none of the three AI assistant crawlers we sent and 4 to only some of them (the study). Often nobody decided that: a firewall or bot-management default did.
- Separate search from training. OAI-SearchBot, Claude-SearchBot and PerplexityBot decide whether you can appear in those engines’ answers; GPTBot, ClaudeBot and Google-Extended govern training. Blocking a search crawler removes you from answers and protects nothing (the details).
- Check your firewall, not just robots.txt. A bot challenge returns a page with no content, and the crawler moves on (diagnosing 403s on Cloudflare).
- Put the answer in the HTML. Several AI fetchers run no JavaScript, so text that appears only after scripts run may be invisible to them (how to test it).
The free AI visibility checker shows in a minute which crawlers can read your home page.
2. Map the sources AI cites for your topics
With web search on, an engine reads a handful of pages and leans on what they say. Those pages are your real competition for the answer, and most of them are not yours: Muck Rack found 84% of the links ChatGPT, Claude and Gemini cite go to earned media (Muck Rack, May 2026).
From your baseline, list every domain the answers cite on your topics and how often. For each one, note whether it mentions you, mentions you wrongly, or leaves you out. In our experiments of 1 October 2026, a brand named on six or more of the pages other answers cited was named in 34.9% of answers, against 11.9% for a brand on none of them (method). That gap is what the next move closes.
3. Earn a place on the pages engines read
This is the slowest move and usually the largest. For commercial questions, “best X” lists are the pages engines cite most: they were 43.8% of cited page types in Ahrefs’ study of ChatGPT answers (Ahrefs), and listicles took about 41% of citations on commercial prompts in Wix’s study of ChatGPT, AI Mode and Perplexity (Wix Studio).
- Get into the lists that are cited, with facts their authors can check: who you are for, what you cost, how you differ.
- Correct what is wrong. An outdated price or a discontinued feature on a cited page will be repeated in answers.
- Be where buyers compare: review sites, communities and forums your category uses, with real customers, not planted posts.
- Publish something worth citing. Original data with a clear method gets quoted by the people whose pages engines read (publishing original research).
Buying mentions or placing fake reviews is not a shortcut: Google’s spam policies cover attempts to manipulate AI answers since May 2026, and an engine that finds a brand only on paid pages has little reason to trust it.
4. Answer the shortlist questions on your own site
When engines search, the words they add show what they are looking for: in our run, 45% of their searches contained “best”, 11% “vs” and 10% “alternative”. Make sure your own site answers those questions plainly:
- Comparison pages with named competitors, honest about where each fits better.
- Alternatives pages for the competitors buyers move from.
- Pricing with its conditions, so a passage can be quoted without guessing (pricing pages for AI answers).
- Who you are best for, and integrations, in plain lists.
Your own pages are rarely the only source an engine reads, but they set the facts other pages repeat. Ahrefs found a brand’s own “best” list still appeared in more than a third of the software answers in which that brand was recommended.
5. Write passages that can be quoted alone
Engines retrieve passages, not whole pages. A paragraph that names its subject and gives the answer in its first sentence can be lifted into an answer; one that relies on “it” and “as mentioned above” carries less (the evidence).
- Lead each section with the answer, then the detail.
- Use specific numbers with their date and source.
- Name the product, the plan or the market in the sentence that makes the claim.
The research that named generative engine optimization found that adding citations, quotations and statistics raised a source’s visibility in generated answers by 30 to 40%, while keyword stuffing made it worse (the details). Treat the size as direction: the benchmark used a 2023 model.
6. Make your facts the same everywhere
Engines cross-check. If your pricing, headcount, positioning or founding date differ between your site, your social profiles, directories and review sites, they hedge or choose the competitor whose facts agree. Give each important entity one defining page and bring the rest in line with it (the evidence, brand entity consistency). If an answer describes you wrongly, trace the claim to the page it came from and fix it there (when AI gets your brand wrong).
7. Change one thing at a time, and judge it
Answers move on their own as engines update and competitors publish. A chart of your share before and after a change will flatter almost any change. The fix is a comparison: aim each change at named topics, and compare those topics with topics you left alone over the same weeks. If the gap is clear, scale the change. If you cannot tell, say so, and wait before doing more of it.
This is what Vercatus does with every change you ship: a verdict against topics left alone, read on days 14, 30 and 45. How it works.
What does not work
- llms.txt and special markup. Google says Search does not use llms.txt and needs no special schema for its AI features (the details).
- A page for every phrasing. Google treats mass-produced pages as scaled content abuse, including separate pages for fan-out queries (the details).
- Keyword stuffing. It lowered visibility in the GEO research.
- Chasing one prompt. Answers to the same question vary; work on topics, measured across many questions.
Questions people ask
How long does it take to improve AI visibility?
Engines that search the web can reflect a change within days of reading the new page, but expect a clear signal in two to six weeks, and judge it against topics you did not change. What a model learned in training shifts only when it is retrained, so being well covered across many independent sources is the long game.
What is the fastest way to appear in ChatGPT answers?
Fix access first: make sure OAI-SearchBot can read your site. Then get onto the pages ChatGPT already cites for your topics, such as the lists and reviews in your category, and answer the comparison and alternative questions plainly on your own site.
Does SEO help AI visibility?
Yes. Engines that search often find pages through search: in our tests, 66.3% of the brands an answer recommended ranked in the top 10 for one of the engine’s own searches, against 30.7% of the brands it left out. SEO is the foundation; the rest of the moves here build on it. GEO vs AEO vs SEO explains how the two relate.
Do I need to publish more content?
Usually not more, but better placed. Most of the effect comes from pages other people publish and from making your existing pages answer the shortlist questions plainly. Mass-producing pages can hurt.
How do I know which change worked?
Compare the topics a change aimed at with topics you left alone, over the same weeks, with enough answers behind each figure. Without that comparison, any before-and-after chart is mostly noise.
Part of our guide to AI visibility: how it is tracked, scored and audited.