How to See Whether AI Recommends Your Brand: a Real Audit

Most brands know what their Google Analytics says about organic traffic. Very few know what ChatGPT says about them.

That is a problem that compounds quietly. A growing number of buyers now start their product research in an AI chatbot rather than a search engine. They type "best greens powder for energy" or "which collagen supplement should I take" and get a confident recommendation: three to five brand names, with brief reasons attached. No links to sort through. No ads to scroll past. Just a shortlist.

If your brand is not on that shortlist, the buyer never sees you. They move to product pages, reviews, and purchase consideration for brands they already encountered in the AI response. Your paid ads, your SEO, and your conversion rate work on a smaller pool of people.

A 2026 consumer survey by PartnerCentric (n=1,004) found that 64 percent of consumers plan to use AI chatbots for shopping in 2026, with nearly one in four intending to make AI shopping their default purchase method. Health and supplements ranked as the fourth most common category for AI-assisted shopping at 41 percent.

The audit described in this piece takes about two hours to run manually. By the end of it, you will know your share of voice in AI-generated category answers, which competitors are dominating those answers, and which source gaps are keeping your brand out of the conversation.

What This Audit Actually Measures

AI visibility is not SEO. They overlap in some inputs but measure different things.

SEO measures where your pages rank in search results. AI visibility measures how often your brand appears in AI-generated answers to buyer questions, and what position it occupies when it does appear. A brand can rank first organically and still be absent from ChatGPT's recommendations. BrightEdge research found that only 16.7 percent of sources cited in Google AI Overviews overlap with the first page of organic search results.

The audit has three distinct measurements. First, mention frequency: across your prompt set, how often does your brand appear in a response? Second, mention position: when your brand appears, is it the first brand named, the third, or buried in a qualifier? Third, citation sources: which external pages is the AI pulling from when it includes or excludes you? Each of these tells you something different and points to a different set of actions.

Running all three gives you a snapshot of exactly where you stand. Most brands start this process with no idea whether they have a problem.

The Five AI Engines to Cover

Different AI engines pull from different source pools and weigh signals differently. Running your audit on only one engine gives you an incomplete picture.

The minimum viable set for a 2026 audit covers five platforms.

ChatGPT (with web browsing enabled): The largest AI by weekly active users, reaching 900 million weekly users as of early 2026. Browsing mode pulls live citations, which is what matters for the audit. Run prompts with browsing on.

Perplexity: Heavy reliance on Reddit and real-time web results. It surfaces different sources than ChatGPT and often reflects community discussion more directly.

Gemini: Google's AI. Research from Surfeo's 2026 audit of 518 businesses found it filters brands below 3.5 stars on review platforms. Strong alignment with Google's editorial trust signals.

Google AI Overviews: Appears in Google search results for broad category queries. BrightEdge found Google AI Overviews are present in 48 percent of all tracked search queries. Different mechanism than standalone chatbot queries but critical for discovery at scale.

Google AI Mode: The newer conversational layer in Google Search. Still emerging, but tracking it now builds your baseline before it scales further.

Brands that appear consistently across AI recommendations tend to have citation infrastructure deep enough to register on all five engines simultaneously. You are looking for both where you stand now and how consistent that presence is across platforms.

Step 1: Build Your Prompt Set

The prompt set is the foundation of the audit. A weak prompt set produces incomplete data. A well-built one surfaces exactly where you stand and which specific buyer scenarios you are missing.

Write 20 to 30 prompts minimum. For a thorough audit, 50 to 60 prompts give you higher confidence. Build them in three types.

Definitional: What is it, and what does it do? These capture how the AI understands your category. Example: "what is a greens powder supplement" or "how does collagen help skin."

Comparison and recommendation: These are the buyer-decision prompts. "Best greens powder for energy," "top rated collagen supplements for women over 40," "which protein powder is best for people who don't like protein taste." These are the most important prompts in the set. They directly reveal which brands the AI recommends at the moment of decision.

Specific use case and problem-based: These mimic the real language buyers use when they have a specific need. "Greens powder that doesn't taste terrible," "collagen supplement that mixes well with coffee," "supplement for someone who hates swallowing pills." These prompts often surface brands with strong niche community presence that broader queries miss.

Write prompts in natural language, not keyword form. "Best collagen supplement" is a keyword. "I'm looking for a collagen supplement my doctor won't roll their eyes at, any recommendations?" is how buyers actually ask AI. Both matter. The conversational form reveals a different signal profile.

Do not reuse prompts from existing keyword research. Your keyword list was built for an algorithm that ranks links. These prompts need to reflect how a real buyer talks to a conversation partner, which is a different register entirely.

One useful exercise: open ChatGPT and Perplexity and type your category name. Look at the suggested follow-up prompts that appear. Those are real queries. Add the most relevant ones to your set.

Step 2: Run the Prompts and Log What Happens

Open a fresh session in each AI engine before each prompt batch. Context carryover from a previous conversation can affect what the model surfaces. You want each session to reflect the model's baseline, not a session already warmed on your category.

For each prompt, log these five fields in a spreadsheet.

The exact prompt as written. Copy it verbatim. You will need to run the same prompts again in 90 days, and prompts that are paraphrased produce different results.

The AI engine and the date. Different engines on the same day can produce different results. Date-stamping lets you track shifts over time.

Whether your brand appeared. Yes, no, or mentioned only in a negative comparison. The third case matters: appearing as a negative reference is better than not appearing, but it is a different signal than a positive recommendation.

Position if you appeared. The first brand named in an AI response receives the most consideration. Third position or later functions more like a mention than a recommendation.

Which brands appeared ahead of you. Or in your place, if you were absent. These are the competitors whose citation infrastructure you will reverse-engineer in the next step.

Also paste the full first response for any prompt where you appeared or where a competitor you care about appeared. You will need that text when you trace the citations.

Here is the structure of what your tracking log should look like:

Fill this in for every prompt and engine combination. Do not skip prompts where you do appear. You need the complete data set, not just the gaps. A session covering 30 prompts across five engines produces 150 data rows. At a comfortable pace, this takes two to three hours total.

Step 3: Map the Cited Sources

This is the step most brands skip. It is also where the most actionable information lives.

For every AI response that mentions a competitor in a position above yours (or in your place when you are absent), look at the sources the AI cited. In ChatGPT with browsing on, these appear as inline footnotes or linked sources at the end of the response. In Perplexity, they appear as a source list on the right side of the screen. In Google AI Overviews, they are the linked cards below the summary.

For each competitor that appeared ahead of you, record three things.

Which domain appeared as the citation source. Write down the full URL. You want to know whether it was a specific Healthline article, a specific Reddit thread, or a specific comparison post, not just the domain name.

The content type. Is it a comparison article, a top-10 roundup, a Reddit thread, a Wikipedia entry, an editorial review, or a brand-owned page? The type matters because each carries different AI weight and requires a different action to match.

Whether that source mentioned your brand. This is the critical question. If the Healthline article that ranks as a top ChatGPT citation covers your top two competitors and not you, that is a specific, addressable gap.

After completing this step across your full prompt set, you will have a clear map of which sources are driving your competitors' appearances. It is almost always a cluster: two or three editorial publications, a handful of Reddit threads, and in some categories, a Wikipedia presence.

The gap between the sources driving competitor citations and the sources that could drive yours is your action plan. Not a vague goal to "build more content." A specific list of articles to get included in, communities to earn a presence in, and editorial relationships to develop.

BrightEdge found that Google AI Overviews appear in 48 percent of all tracked search queries. That means nearly half of all category-level Google searches return an AI-generated answer drawing from editorial and community sources. Those sources are the territory worth mapping.

Step 4: Score Your Share of Voice

Once your log is complete, score it.

Share of voice for AI visibility is simple arithmetic. Take the number of prompts where your brand appeared (in any position) divided by the total number of prompts you ran on each engine. That percentage is your mention rate per engine.

Then calculate position share: of the prompts where you appeared, how often were you the first brand named? Second? Third or later? Position matters because the first brand named in an AI response receives the most consideration. Third position or later is more a mention than a recommendation.

Track both numbers. A brand with a 60 percent mention rate but consistently listed third has a different problem than a brand with a 40 percent mention rate where it consistently appears first.

Re-run this audit every 60 to 90 days. The citation landscape shifts with new editorial content, algorithm updates, and changes in what each platform treats as authoritative. Your first audit is the baseline. The second tells you whether the actions you took are registering.

Tools like Ahrefs Brand Radar, Profound, and Brandwatch can automate this tracking across multiple engines continuously. For the manual audit, the spreadsheet approach above is enough to establish a useful baseline in a single session.

What the Data Typically Shows

The Surfeo analysis audited 518 businesses across 20 countries between February and April 2026, testing visibility across ChatGPT, Gemini, Perplexity, and Claude. The headline finding: 91 percent of businesses scored below the threshold for meaningful AI visibility, with an average score of 38 out of 100. More than half (56.6 percent) scored below 40.

This is not because most brands have bad products. AI citations are concentrated among the few brands that have built specific infrastructure. The Surfeo analysis found that appearing on sector listicles and comparison articles accounts for 41 to 64 percent of a brand's recommendation weight in AI responses.

Most brands that run this audit for the first time find one of three patterns.

Invisible: Absent from nearly all prompts across all engines. This is the most common result by a wide margin. It means the editorial and community infrastructure needed for AI citation simply does not exist yet.

Inconsistently present: Appearing in some prompts but not others, with no clear pattern. This usually indicates narrow editorial coverage, typically limited to one or two publications that have cited the brand.

Present but outranked: Appearing in most prompts but consistently listed behind one or two competitors. This indicates those competitors have deeper presence in the specific sources the AI trusts most, often comparison roundups on publications with high editorial authority.

Because the Ahrefs Brand Radar add-on was not active for this audit run, the worked example above is structural, showing the format and process rather than live figures for a specific category. The methodology applies identically whether you run it manually or through an automated platform.

The Four Signals to Look for in Your Results

Once you have your log and your source map, examine four specific signals.

Editorial coverage on trusted domains. Which publications appear repeatedly as citation sources for your top competitors? If Healthline, Wirecutter, a category-specific editorial outlet, or a major media property keeps showing up, those are the editorial relationships worth pursuing. Surfeo's 2026 data found that off-site brand mentions carry three times the AI visibility impact of classic backlinks. Being included in a trusted editorial roundup carries more weight than building links to your product page.

Community presence, especially on Reddit. How many organic Reddit threads discuss your brand in substantive terms? Competitors with consistent AI citations often have active communities discussing their product across multiple relevant subreddits. Perplexity draws heavily from Reddit: research finds it accounts for up to 46.7 percent of Perplexity's top citations. A brand with minimal organic Reddit discussion is missing Perplexity's primary citation source.

Review density and recency. Surfeo found that Gemini filters businesses below 3.5 stars. Beyond that threshold, recency matters: AI platforms pull from what is current, not what was true two years ago. Check the date distribution of your reviews on Google, Trustpilot, and relevant category platforms. If most of your reviews are older than 12 months, you are likely losing citation weight to brands with fresher review activity.

Whether editorial coverage of your brand contains external citations. Surfeo research found that content with external citations increases AI visibility by 115 percent compared to content without them. Articles that mention your brand and include data with sourced statistics are significantly more likely to be pulled by AI than articles that mention you in passing. When you earn editorial placements, push for inclusion in data-rich pieces with sourcing.

How to Act on What You Find

Your source map is your action list.

The fastest lever is outreach to existing editorial content that is already being cited by AI. Comparison articles and top-10 roundups that rank well and appear repeatedly in your citation map are the targets. Getting added to five of those articles shifts your citation footprint faster than building new content. Contact the publication, lead with a specific and compelling angle about your product, and aim for inclusion in existing content rather than commissioning new pieces that need time to build authority.

For brands with no Reddit presence: this cannot be manufactured quickly. Build the product and community experience that generates genuine conversation. Founder AMAs in relevant subreddits, honest outreach to category reviewers, and customer community programs take 6 to 12 months to compound into meaningful citation signal. Start now rather than next quarter.

Review velocity matters more than your total review count. A brand generating 20 new reviews per month has fresher citation infrastructure than a brand that accumulated 500 reviews two years ago and slowed down. Build the review cadence into your post-purchase email flow.

For editorial coverage that is more than 12 months old: reach back to publications that featured you previously and propose updated comparisons. A refreshed placement re-enters the AI citation pool quickly. Content freshness is a short-cycle variable you can move without starting from zero.

For more on the specific citation signals that drive AI recommendations, see our piece on what actually decides which brands ChatGPT recommends and how to get your brand mentioned by ChatGPT, Gemini, and Perplexity. For the divergence between AI citations and organic rankings, this piece covers why the two engines reward different things.

Frequently Asked Questions About the AI Visibility Audit

If you want a read on where your brand stands in AI recommendations and what it would take to build consistent visibility, we run this audit as part of how we evaluate every new account.