How brands show up in AI answers, and what to measure
The same question rarely gives the same answer twice, so one AI answer says little about your brand. How often you appear, across many runs, says more.
By Nathan Barling Product Marketing Principal 4 min read
Ask an AI assistant the same question many times and the brands it names will change. In SparkToro's study, the same prompt returned the same list of brands less than 1% of the time.
One AI answer tells you almost nothing about your brand. What tells you something is how often you appear across many runs of the same questions, and whether you're named, linked, or both. Those are two different things to count.
Figure 1
AI engines treat brands differently: some name them, some link them
Share of brand appearances, by engine
Bar chart: ChatGPT cites brands 87% and mentions them 20.7%, and Gemini cites 21.4% and mentions 83.7%
One answer is a sample of one
SparkToro, working with the AI tracking vendor Gumshoe.ai, ran 2,961 prompts across three AI engines, including ChatGPT and Google AI Overviews, with each prompt run 60 to 100 times per platform. ChatGPT and AI Overviews each returned the same brand list less than 1% of the time. The same list in the same order came back less than 0.1% of the time.
Frequency held up better than rank. In a headphone test, four well-known brands appeared in 55% to 77% of responses. So the useful number is how often you show up, not where you sat in one answer.
The study isn't peer reviewed, and one of the two firms behind it sells tracking. Even so, the direction matches what anyone sees who asks the same question twice.
Being named and being linked are different
A brand can appear in an AI answer in two ways. It can be named in the text, which is a mention. Or its website can be linked as a source, which is a citation. Semrush found these split by engine. ChatGPT cited brands 87% of the time but named them in only 20.7% of answers. Gemini did the reverse: named 83.7% of the time, linked 21.4%.
Across all four engines in the study, 74.9% of brand appearances included a citation and only 38.3% included a mention. A page can be cited constantly while the brand name never appears. Or a brand can be recommended without sending a visit. Track one and you will misread your position.
How much traffic it sends
Usage is broad. Pew found that 49% of US adults use AI chatbots, and 42% use them to search for information.
Referrals are small so far. SE Ranking found ChatGPT accounted for 0.32% of website referral traffic in May 2026, across 101,574 sites, and we found no consistent evidence on how well those visits convert. So treat AI visibility as a brand and consideration measure first, and count the visits as a bonus.
What you can control
Start with what the engines themselves say. Google says its generative AI features are rooted in its core ranking and quality systems, and that you don't need new machine-readable files, AI text files, or special markup. Structured data isn't required. llms.txt is a proposed format and isn't an adopted standard yet.
OpenAI's documentation says sites opted out of OAI-SearchBot won't be shown in ChatGPT search answers. That's a statement about crawler access. We found no OpenAI statement on how content is chosen for citation.
The work that follows is mostly ordinary. Let the crawlers in. Answer real questions clearly, with definitions a reader could quote. Keep facts current and dated. That's the base of generative engine optimization and answer engine optimization, and it's also what makes a page worth citing.
What to measure
Pick a fixed set of questions and run them on a schedule. Then report these four, by engine, so you can see a share of LLM presence that holds still long enough to compare.
Presence
Share of runs where your brand is named, across a fixed prompt set, by engine.
Citations
How often your domain is linked, and which of your pages are used.
Accuracy
Whether the description of your brand is correct and current.
Traffic
Visits from assistants, tracked as their own channel.
A composite such as an AI visibility score can be handy for reporting. Keep the parts visible underneath it, since a mention and a citation can move in opposite directions. It also belongs next to your other brand tracking, not in place of it.
Where Alli fits
PMG's teams, including its Forward Deployed Engineering team, do GEO, AEO, and llms.txt work for clients. Alli Brand Equity measures whether it's working. Its Share of LLM Presence measure sits alongside the brand's other signals, so teams can see how often the brand appears in AI answers next to the reputation signals that shape customer choice. It's one input to planning, and a ranking to watch rather than chase.
How much to trust this
Nearly every dataset on AI visibility comes from a company that sells AI visibility or SEO tools: Semrush, SE Ranking, SparkToro, and Gumshoe.ai. Pew, Google, and OpenAI are independent of that market but say less. Prompt sets are small, and vendors define a mention differently. Hold conclusions loosely and re-check quarterly.
Run your 25 questions as many times as you can afford, on the engines your buyers use. That first honest baseline beats any benchmark in this article.
Frequently asked questions
What is share of LLM presence?
Share of LLM presence is the proportion of AI answers to a set of prompts that name or cite your brand. Measure it across many runs, because the same prompt rarely returns the same list. It's a frequency, not a rank.
What is the difference between a mention and a citation?
A mention is when an AI answer names your brand in its text. A citation is when it links your website as a source. Semrush found ChatGPT cites brands far more than it names them, and Gemini does the opposite.
What is generative engine optimization?
Generative engine optimization is shaping content and brand signals so AI models can find and cite your brand. In practice it overlaps heavily with clear, current, well-sourced content.
Do I need schema, llms.txt, or special files to appear in AI answers?
For Google, no. Google says you don't need new machine-readable files, AI text files, or special markup, and structured data isn't required. llms.txt is a proposal, not an adopted standard. For ChatGPT search, OpenAI says sites that block OAI-SearchBot aren't shown.
How often should I measure AI search visibility?
Monthly is a reasonable start. Use the same prompts and engines each time, and run each prompt many times. SparkToro found identical lists appear under 1% of the time, so a single check tells you little.
Sources
- Pew Research Center, Americans and AI 2026: Chatbots, Smart Devices and Views on Impact, June 17, 2026 (n=5,119).
- Semrush, The Ghost Citations Study, June 9, 2026 (3,981 domain appearances, 115 prompts, 14 countries, four engines). Semrush sells AI visibility tools.
- Search Engine Journal, AI recommendations change with nearly every query, January 30, 2026, reporting SparkToro research with Gumshoe.ai. Both sell related tools; not peer reviewed.
- SE Ranking, ChatGPT referral traffic, May 2026, July 9, 2026. SE Ranking sells SEO software.
- Google Search Central, Optimizing your website for generative AI features on Google Search, updated July 10, 2026.
- OpenAI, Overview of OpenAI Crawlers. Undated.
Last updated November 6, 2026. Statistics are quoted from the sources above and were checked on that date. Survey results are self-reported unless noted.