GEO: How to Get Your Company Cited in AI Answers

Being cited means your page is used as a source, linked below or beside the answer. Being named means your brand is written into the text the user reads. Semrush separated the two outcomes across 3,981 domain appearances, collected in 14 countries and four engines: ChatGPT, Google AI Overviews, Google AI Mode and Gemini¹.

Every appearance fell into one of three buckets:

• Ghost citation, with a link and no brand name: 61.7% of cases¹.
• Citation with the brand named in the text: 13.2%¹.
• Brand named with no link to any of its pages: 25.1%¹.

Across the buckets, 74.9% of appearances included a citation, but only 38.3% included the brand name¹. For anyone accountable for demand generation, the gap matters: a citation credits your content and may earn a click; a mention is what sticks with the reader who never clicks.

Each AI engine leans one way, and visibility in one does not carry over to another. In the Semrush study, ChatGPT cited the brand in 87% of appearances and named it in only 20.7%¹. Gemini did the reverse: it named the brand in 83.7% and cited it in 21.4%¹. Google's AI Overviews and AI Mode landed in between, closer to citing.

BrightEdge looked only at ChatGPT, with a different method, and reached what looks like the opposite result: ChatGPT mentions brands 3.2x more than it cites sources, averaging 0.74 citations per prompt, and 44% of prompts include no brand at all². The gap comes from the unit of measurement. BrightEdge counted every brand in each answer, on queries with heavy purchase intent; Semrush recorded, for each domain that appeared, whether it was cited, named or both.

The practical takeaway is the same in both studies: tracking only "did we show up as a source" hides half the picture. Both are observational studies with no control group, so they show patterns, not causation.

Brazil sits on both sides of the problem. AI adoption is high: Bain's Consumer Pulse 2026 survey, with 2,000 Brazilian respondents, found that 77% of Brazilians use artificial intelligence, and 79% name ChatGPT as a tool they use³. At the same time, AI answers in the country rarely name the brands they use as sources.

The usage numbers, all from the same Bain survey:

• 65% use AI for general research³.
• 43% use AI to look up product or service reviews³.

In the Semrush study, Brazil, Italy and the Netherlands had a brand named in only 18% to 22% of answers, even though the engines cited sources from those markets 82% to 94% of the time¹. India and Sweden reached 50%¹. Read this as a signal, not a rule: the sample is 115 prompts in total, split across 14 countries¹. For a company selling into Brazil, the reading is simple: being a source is likely; being named takes deliberate work.

Ranking well on Google helps, but it is no longer enough. In the 2026 update of the Ahrefs study, covering 863,000 keywords and 4 million cited URLs, only 37.9% of AI Overview URLs appeared in the top results blocks⁴. In July 2025, the same analysis had found about 76%⁴.

The rest split almost evenly: 31.2% of cited URLs ranked in positions 11 to 100, and 31.0% did not appear in the top hundred at all⁴. Ahrefs attributes the shift to query fan-out: the system breaks the question into several sub-queries and cites pages that rank for those, not just for the original query. Google's documentation on AI features confirms that AI Overviews and AI Mode use this technique. YouTube alone accounted for 5.6% of all cited URLs⁴.

With AI assistants, the gap from rankings is wider. Across 15,000 prompts, only 12% of links cited by ChatGPT, Gemini and Copilot were in Google's top 10 for the same question, and 80% of citations came from pages that did not rank for it at all⁵. Perplexity was the exception, with 28.6% of its citations in the top 10⁵.

Mentions depend less on your website and more on what the web says about your brand. Ahrefs matched 75,000 brands against millions of AI Overview responses and found the strongest correlation, 0.664, between branded mentions on other sites and mentions in the answers⁶. Branded organic traffic correlated at only 0.274⁶.

The effect is concentrated. Brands in the top quartile for web mentions earned over 10x more AI Overview mentions than brands in the next quartile down⁶. Ahrefs repeats the caveat that applies to this whole article: this is correlation, not causation. Well-known brands get mentioned more across the web and also get named more by AI, and the study cannot separate the two. In the expanded study covering ChatGPT and AI Mode, YouTube mentions took the top spot as the strongest correlating factor on all three systems.

Query type changes the outcome too. In the Semrush data, informational queries ("what is", "how does") had an 89.3% citation rate and only an 18% mention rate¹. Comparative queries ("best", "vs", "recommend") reached a 43.3% mention rate, 2.4x more than informational ones¹. Explanatory content becomes raw material; comparisons force the AI to name names.

Google says there are no extra technical requirements to appear in its AI features: the page has to be indexed and eligible to show in Search with a snippet. It also does not require a new AI text file or special structured data markup. The work, then, is in your content and your brand's footprint, not in a technical trick.

What we recommend, in order of effort:

1. 01
   Allow crawling. Check that robots.txt and your CDN do not block Googlebot or OAI-SearchBot, OpenAI's search crawler, and that no key page uses nosnippet.
2. 02
   Answer first. Open every page with the direct answer to the question and put the sourced data point right after it; that passage is what AI tends to reuse.
3. 03
   Cover the sub-questions. For each topic, list the adjacent questions (cost, timeline, comparison, risks) and answer all of them on the same page or on pages connected by internal links, because query fan-out searches for exactly those variations.
4. 04
   Put your brand name next to what is yours. Proprietary data, methods and case studies should carry your name in the text ("in Company X's survey"), so an AI that uses the passage has the name at hand.
5. 05
   Publish honest comparisons in your category, with clear criteria and real competitors. That is the format that earned the most mentions in the Semrush study.
6. 06
   Build presence off your site: trade press, interviews, YouTube videos, industry forums and directories. It is the factor most tied to mentions in the Ahrefs data.

The measurement we recommend is a fixed panel of questions from your category, run on the same engines every month, logging citations and mentions separately. AI visibility tools such as Semrush's AI Visibility Toolkit and Ahrefs Brand Radar automate this, but they measure their own prompt panels, not your customers' real questions, so it pays to keep your own.

1. 01
   Build the question list from what your sales team hears and from your Search Console queries, split by intent: informational, comparative and commercial. Write each one in a short and a long version, because prompt length changes the mention rate.
2. 02
   Run every question in ChatGPT with search on, in AI Overviews, in AI Mode and in Gemini, in a session with no history and with the location set to your target market.
3. 03
   For each answer, log three fields: was your domain cited (yes or no), was your brand named (yes or no), and which competitors were named.
4. 04
   Calculate the citation rate (answers with your domain among the sources divided by the total) and the mention rate (answers with your brand in the text divided by the total), by engine and by intent.
5. 05
   Add up share of mention for the category: of all brands named across the panel, how often it was yours versus each competitor.
6. 06
   Tie it to the business. In Google Analytics, build a segment for referral traffic from chatgpt.com, gemini.google.com and perplexity.ai, and track sessions, conversions and revenue from that channel next to the panel rates.
7. 07
   To separate correlation from incrementality, change one group of pages first and leave a similar group untouched. If rates and conversions rise in the changed group and not in the control group, the effect is yours; if they rise together, it was the market.

AI answers vary from one run to the next, so a single answer proves nothing. Watch the panel trend over months, not one day's result.

The first deliverable is a spreadsheet with the questions your buyer would ask before purchasing, split by intent, plus columns for citation, mention and competitors named. Run the first round on all four engines and log the results: that is your baseline, and every content change gets compared against it.

With the baseline in place, start with the comparative and commercial questions where a competitor is named and your brand is not. Those are the ones that call for a comparison page, a signed data point and off-site presence, in that order.