Getting AI to recommend your business depends on how often credible sources name your brand next to the problem a buyer asks about, and how clearly they describe it. Assistants such as ChatGPT and Gemini build their shortlists from what independent sources repeat in press coverage and reviews, as well as on forums and reference pages. Your own site states the facts and third parties confirm them. The brands that come up most are the ones the web already discusses in the context of the buyer's question.

AI picks the brands it has seen named most often by credible sources in connection with the topic of the question. Ahrefs studied 75,000 brands and found that branded web mentions had a 0.664 correlation with visibility in AI Overviews. That's stronger than any backlink or domain metric it measured. Brands in the highest quartile for mentions averaged 169 AI Overview mentions, while the next quartile averaged 14.
AI doesn't keep a fixed ranking. SparkToro ran 2,961 tests in which volunteers repeated the same prompts. They found a less than 1-in-100 chance that ChatGPT or Google's AI would return the same brand list twice. Rand Fishkin, SparkToro's CEO, said that "any tool that gives a 'ranking position in AI' is full of baloney." The pool of candidates stayed steady, though. In the headphone prompts, Bose and Sony each showed up in 55% to 77% of responses, as did Sennheiser and Apple.
So aim to be part of that steady pool. Before you change anything, check whether people who write about your category name you when they list the options.
AI assistants draw their answers from 2 sources, their training data and pages fetched through live web searches. Training data shapes which brands the model links to a category in the first place. Live retrieval adds current pages, which is why a strong review published last month can show up in answers before the next model is trained.
Google describes AI Overviews and AI Mode as using a "query fan-out" technique that involves issuing "multiple related searches" across subtopics before they write the response. A buyer who asks for the best payroll software for a 20-person firm sets off hidden searches about pricing and integrations. So the assistant reads your pricing page even when the buyer never asked about price.
ChatGPT decides for itself when to search the web, while Perplexity searches on every question. To see which mode you're up against, run your key prompts in each assistant and look at whether the answer lists sources.
AI models trust signals that come from outside your own site. The most important include how often you're mentioned and how consistently you're named. How much people search for you also matters. In its larger study, Ahrefs compared those signals across ChatGPT and AI Mode. The study also covered AI Overviews. Domain strength and backlinks, the signals classic SEO is built around, came last.
| Signal | Correlation with AI visibility | What it measures |
|---|---|---|
| YouTube mentions | 0.737 | Your brand named in video titles, descriptions and transcripts |
| Branded web mentions | 0.656 to 0.709 | Your name on crawlable pages outside your own site |
| Branded anchor text | 0.511 to 0.628 | Links that use your brand name as the clickable text |
| Branded search demand | 0.352 to 0.466 | How often people search for your name |
| Domain Rating | 0.266 to 0.326 | Overall backlink strength of your domain |
| Backlinks | 0.218 | Raw count of links pointing to you |
YouTube mentions
Branded web mentions
Branded anchor text
Branded search demand
Domain Rating
Backlinks
*Spearman correlations from Ahrefs' 75,000-brand study. Ranges cover ChatGPT as well as Google's AI Mode and AI Overviews.*
Correlation doesn't prove cause. Brands people discuss are also brands people search for. Still, the order tells you where to spend effort first.
Researchers from Princeton and partner universities tested rewrites of web pages and found that sourced statistics and quotations, when added alongside citations, raised a page's visibility in generative answers by up to 40%. The effect changed by subject area. Keyword stuffing didn't help.
Structured data on its own doesn't raise how often AI cites you. Google's documentation states, "Structured data isn't required for generative AI search, and there's no special schema.org markup you need to add." Google still recommends schema for ordinary SEO, because it makes pages eligible for rich results.
Independent data points the same way. In 2026, Ahrefs tracked 1,885 pages that added JSON-LD against about 4,000 control pages that didn't. The changes in citations from ChatGPT and AI Mode couldn't be told apart from noise. Citations from AI Overviews moved slightly down. A companion test found that the AI systems it checked read only the visible HTML when they fetched a page directly.
Keep your Organization and Product markup, since Google uses it to understand who you are. Then check that it matches the visible page word for word. Facts a reader can see carry more weight than hidden code.
The gaps worth going after are the prompts where assistants name a competitor and leave you out, and the sources behind those answers. You can find both with a manual audit before you pay for any tool.
Close source gaps first. When the same review site or comparison article keeps feeding answers that leave you out, getting listed there changes what the assistant reads. Content gaps come next. These are pages you've never written, like an honest comparison against your main rival.
Comparison prompts are the most open ground. In a Seer Interactive study, AI engines declined to answer 80% of comparison prompts about Seer's own brand. When assistants have nothing to say, the first brand to publish a clear, factual comparison gives them their material.
Hallucinations put your brand at risk because a buyer rules you out over a fact that was never true, and nobody tells you it happened. Wrong prices and features from discontinued products are the classic cases. An assistant that quotes 3 times your real price ends the evaluation before the buyer ever reaches your site.
The error rate is lower than the fear suggests, but it isn't zero. Seer Interactive tested 28,123 AI responses about its own agency. When the platforms answered, they were accurate 95.9% of the time. Inaccurate answers averaged 2.8% and reached 5.7% in AI Overviews. More often, the assistants simply declined to answer, which happened on 31.2% of prompts.
Errors come from old or thin sources. You can change what a model reads. Publish 1 plain page with your current pricing and products, and update directory and review profiles that still list old details. Then ask the assistants about your brand again each month to check whether the fixes have reached them.
AI assistants rely mostly on earned and third-party sources, and each assistant has its favorites. Muck Rack's October 2026 "What Is AI Reading?" report looked at 207 million citations across ChatGPT and Claude, with Perplexity and Gemini also covered in the report. It found that more than 84% of citations came from earned and third-party sources such as journalism and academic research. Encyclopedic pages also fell into this category. Paid content made up just 0.3% of citations. Owned media came to 15.2%.
Profound looked at the 10 sources each platform cites most. Wikipedia took 47.9% of ChatGPT's citations within that group. On Perplexity, Reddit held 46.7% of the top-10 share. Gemini leans on YouTube. Business software buyers will also find review platforms such as G2 in ChatGPT's top 10.
So the sources to work on depend on which assistant your buyers use. Look at the sources your audit turned up and start with the domains that appear across more than 1 assistant.
Your AI recommendation share is the percentage of answers to a fixed set of buyer prompts that name your brand. It's a better measure than position, because position changes from run to run. A rate taken across many runs holds steady enough to track month over month.
Repeat each prompt, because single answers drift. Conductor ran a 14,000-call test and found that any 2 runs of the same recommendation prompt shared 49% of brands. Use the same prompts each month and work out the mention rate for you and for each competitor. Keep a separate record of whether each mention was accurate.
Count mentions as well as citations. Semrush analyzed 3,981 domain appearances and found that 61.7% were "ghost citations", where the engine used a page as a source but never named the brand in the answer. A tool that only tracks links will overstate how often buyers actually hear your name.
Start with the audit from the competitor section, then put your time into the channel where your losses cluster. If the answers cite your site with stale facts, fix your pricing and product pages first. If the answers draw on reviews or press coverage that leaves you out, move your effort to earning mentions there. Do the same for forums the answers draw on. Run the same prompts a month later to see whether your share moved. That loop is how to get AI to recommend your business over time.
Snoika is an SEO, Generative Engine Optimization (GEO) and Answer Engine Optimization (AEO) platform that audits websites and tracks brand visibility in Google and AI assistants, with human review on top. To see which prompts name you and which name your competitors, ask Snoika to track your brand's visibility in AI assistants.
AI tends to recommend brands that credible sources repeatedly connect with the buyer's question. The shortlist changes between answers, so appearing consistently across repeated tests matters more than being named first in one response.
Independent brand mentions and consistent naming are important signals. The page's cited research found stronger associations between AI visibility and brand mentions than backlinks. These correlations don't prove that increasing mentions alone will cause more recommendations.
Structured data alone hasn't been shown to increase AI citations. Keep relevant markup for ordinary search features, but make sure it matches the visible page. Put current business facts in text readers and assistants can access.
Test 20 to 40 unbranded buyer questions across several assistants while logged out, repeating them on different days. Record named brands and cited sources, then compare mention rates. Check which repeatedly cited pages include competitors but leave your business out.
There's no fixed timeline for how to get ai to recommend your business. New sources can appear through live searches before a model is retrained. Repeat the same buyer prompts monthly to check whether your mention rate has changed.
Keep reading
All articles


See how AI talks about your brand