The Visibility Question Nobody's Asking
You rank #1 on Google. Congrats. What do ChatGPT, Perplexity, and Gemini say about your brand? Because if you're like most marketers I've talked to lately — the honest ones, at least — the answer is either "no idea" or "not much."
This isn't hypothetical anymore. Generative AI tools have hundreds of millions of users. They're not novelty toys. They're how people discover products, compare services, and form opinions before they ever type a query into a traditional search bar. And the uncomfortable truth? Your SEO ranking tells you almost nothing about whether these systems will recommend you.
I've been digging into how brands measure their footprint across these platforms. Here's what actually works.
Why Ranking First Doesn't Mean Getting Recommended
Traditional SEO measures position on a single results page. AI search works differently. Answer engines don't just index pages and sort them by relevance signals — they synthesize information from multiple sources and decide which brands to mention based on perceived authority, clarity, and relevance to the specific question being asked.
That means a competitor with a mediocre Google ranking could show up in ChatGPT's answer while you don't appear at all. The synthesis process is opaque. Two different people can ask the same question and get different answers. Your position isn't fixed — it shifts based on phrasing, context, and even which model version is running.
According to Semrush's research on AI visibility, the key metrics that matter here aren't rankings. They're mentions, citations, position within the answer, and sentiment. Whether your brand appears at all. Whether the AI links to your site. Where you sit relative to competitors in the response. Whether the description sounds like a fan or a hater.
Building Your Prompt Inventory
You can't track what you haven't defined. The first step is building a prompt list that mirrors how your audience actually asks questions across different stages of their buying journey.
Semrush breaks this into three categories worth stealing:
- Research prompts: "What is X," "how does X work," "best practices for X"
- Comparison prompts: "X vs Y," "best tools for [activity]," "alternatives to Y"
- Evaluation prompts: "Is X worth it," "X pricing," "pros and cons of X"
The trick is specificity. Broad prompts like "best project management software" are useful but generic. Highly specific phrases — the kind your customers actually type into these tools — give you a more actionable picture of your visibility.
Two places to mine for prompt ideas: Google's People Also Ask box for keywords you're already targeting (these often mirror AI platform queries), and Reddit. Search relevant subreddits for your primary keywords, sort by "Top," and the questions that bubble up reflect real audience language.
Manual Tracking Still Works (And You Should Start There)
Before you buy anything, do this manually. Enter each prompt from your list into ChatGPT, Claude, Gemini, Perplexity, whichever platforms your audience actually uses. For each response, log:
- Whether your brand appears in the answer
- Whether the answer includes a link to your site
- Where your brand appears relative to other brands mentioned
- How your brand is described (positive, neutral, negative)
Keep it in a spreadsheet. Columns for the prompt, platform, date, and each data point above. Run through the full list once per week. When you re-run prompts, note what changed since last time.
This gives you a baseline. And honestly, for most small and mid-size brands, that baseline alone is revealing enough to justify action. You'll quickly see which prompts you're missing from and what type of content competitors are getting cited for instead.
Dedicated AI Visibility Tools
Once manual tracking becomes unmanageable, or if you're managing visibility for multiple brands, dedicated tools earn their keep.
HubSpot's AEO tools work by systematically querying AI models, capturing responses, and analyzing how those responses reference your brand, competitors, and industry topics. Their AI visibility tracking submits prompts across ChatGPT, Perplexity, Google AI Overviews, and Gemini, recording whether your brand appears in responses across various prompt variations and intent categories.
Semrush's AI Visibility Toolkit covers similar ground with prompt-level tracking across ChatGPT, Google AI Mode, and Gemini. You get position changes over time, share of voice data, overall sentiment, and topic associations. The key advantage over manual work: surface-level trends you'd miss checking one prompt at a time start showing up when you're querying hundreds.
These tools differ from traditional SEO crawlers in a fundamental way. SEO crawlers scan web pages. AI visibility tools query large language models directly and interpret how those models select sources. Completely different mechanism, completely different data. And as this space matures, it's worth understanding the ethical fault lines of generative engine optimization, because paid mentions and synthetic visibility can undermine exactly the trust you're trying to measure.
Where AI Systems Actually Pull From
This one surprised me. LinkedIn, YouTube, and Reddit are among the most-cited domains in AI-generated answers. Not blog posts. Not your homepage. Community and video content.
That has implications. If you're building an AI visibility strategy and ignoring your Reddit presence or treating YouTube as secondary, you're leaving massive citation opportunities on the table.
The practical move: repurpose existing high-quality content into formats that live on these platforms. Turn a research study into a LinkedIn carousel. Break down a long-form guide into a series of short clips. Use key findings as podcast episodes. AI systems that pull from these surfaces will encounter your brand in contexts you didn't originally control, but you can shape what they find.
Technical Prerequisites You Can't Skip
One finding from Semrush's AI search optimization research that deserves emphasis: if your content never appears as an AI citation source, you might have a crawlability problem masquerading as a content problem.
Check that your most important pages aren't blocked by login walls, paywalls, JavaScript-only navigation that crawlers can't follow, or server errors. Confirm your canonical tags are present and pointing to correct pages. Broken canonicals don't necessarily block AI crawlers, but they can cause AI systems to cite the wrong version of your content or skip it entirely as duplicate.
This is the unglamorous layer. Nobody wants to audit canonical tags when they could be writing content. But I've seen brands pour energy into content optimization while their technical setup quietly prevents any AI system from ever reading the page.
Schema Markup and Structured Summaries
Two content-level tactics that directly improve AI citability:
Schema markup helps search systems map your content to standardized data structures they understand. Article schema, FAQ schema, product schema, these give AI systems unambiguous signals about what your page contains and what type of content it is.
Structured summaries at the top of pages make your content easier for LLMs to scan and cite. This can be a comparison table, a TL;DR block, or a bulleted overview of what's covered. The logic is simple: when an AI system is deciding whether to pull a fact from your page, a scannable summary at the top makes your content lower-effort to process. Lower effort often means chosen. For a deeper look at formatting and hierarchy decisions that make pages easier for AI systems to parse, see our guide to structuring content for AI search.
Specificity Beats Vagueness for Citations
Here's a concrete tactic with outsized impact. Add specific statistics, with sources, to your content. AI systems reference content with specific, sourced data more often than content with vague generalizations.
Replace a sentence like "AI Overviews are appearing in many searches" with something like "As of December 2025, Google's AI Overviews appear in about 15% of search results according to Semrush Sensor." Use the source name, the date range, and the methodology. If you want a model of what a data-led claim looks like in practice, our breakdown of how AI Overviews are reshaping Google search shows how one well-sourced statistic can anchor a whole argument.
This single change on your top-performing pages often produces the fastest citation improvements I've seen documented. Thirty minutes of work per article. Not bad ROI.
The Brand Consistency Problem Nobody Talks About
AI systems build a composite picture of your brand from everything they can access. Your about page, product descriptions, social bios, directory listings, G2 reviews, Reddit threads, all of it gets synthesized.
If these properties describe your brand differently, the AI may construct an inconsistent or inaccurate picture. Audit your homepage, product pages, social profiles, bios, and third-party listings for conflicting language about what you actually do.
Then extend that to active reputation management. Responding to reviews on platforms like G2 and Trustpilot matters. Engaging in relevant Reddit and Quora threads matters. These responses literally become part of the training corpus that AI systems read.
The Tally example from Semrush's research is instructive: their co-founder Marie Martens regularly engages on Reddit, and in every response she reinforces the same core message that appears across all Tally's owned properties, unlimited forms and submissions for free. That consistency compounds. AI systems encountering her responses, the product page, and the about page all get the same signal.
Start Small, Measure Honestly
I'm not going to pretend there's a mature analytics dashboard for this. The tools are new. The measurement frameworks are still stabilizing. But that's not an excuse to do nothing.
Pick ten prompts. Run them through three platforms. Log the results. Check again next week. You'll learn more in an afternoon of manual tracking than you will from reading another conference talk about AI search.
The brands that win here won't be the ones with the most sophisticated tool stack. They'll be the ones who started paying attention to this while it was still early enough to actually move the needle.
