The Inversion: Competitor Visibility Is Good News
Improving visibility in ChatGPT can start by studying why competitors appear in AI-generated recommendations when your brand does not. Matt of Kenyon Digital argues that repeated competitor appearances reveal useful clues about the websites, mentions and content patterns associated with AI visibility. Because the same prompt can produce different answers across runs, the process is not as simple as checking a fixed ranking; competitor analysis instead becomes a way to identify recurring sources and opportunities that may increase a brand's chance of being included.
His case rests on an inversion of how search professionals usually think. SEOs are used to abundant data (Google Search Console, keyword tools, competition scores) while ChatGPT offers none of it: no prompt database, no search console equivalent, no visibility into what people are asking. But there is one saving grace. When ChatGPT answers, especially with web search, it pulls from specific websites, and those sources are visible in the response's source tab. That, he says, is the window.
This is why competitor presence in AI answers should be read as intelligence rather than defeat. Every time a rival shows up, it leaves a trail of breadcrumbs revealing which sources the AI trusts and which content types it prefers. Your competitors, in effect, did the hard work of figuring out what ChatGPT likes; the task is to follow the trail, understand what they are doing right, and do it better.
What AI Models Parse Best On-Site
- Lead with factsAI tends to skip marketing fluff and look for content that directly answers questions, a pricing page buried under paragraphs of brand story may never get read.
- Use semantic HTMLClear H2s, H3s, bullet points, and structured formats let AI parse information far more efficiently.
- Tables winComparison data, features, pricing tiers, and specs belong in tables, a format Matt says AI loves.
- Add FAQ schemaA well-structured FAQ with clear questions and concise answers is among the easiest content for AI to extract and cite.
Running the Prompt Audit
The methodology starts with a test set. Matt recommends writing down 20 to 30 prompts a potential customer might actually type (broad discovery questions, comparisons, and a few naming competitors directly) and drawing phrasing from Google's people-also-ask boxes and Reddit threads in the niche. AI assistants themselves, he notes, are good at helping brainstorm this list.
Then run the prompts manually in an incognito window, logged out, so personalization does not skew results. Since ChatGPT processes roughly 2.5 billion prompts a day with 800 million to a billion weekly active users, it is a sensible starting platform. The snag is that there is no easy way to log the sources it displays, so Matt suggests a workaround: after each response, type "Please format this response as markdown with all the source URLs explicitly included, " then paste everything into a spreadsheet, one row per prompt.
The analysis phase is a hunt for overlap: which competitors appear most often, on which platforms, for which prompt types, and above all, which sources they keep appearing in. The same types of sources recur again and again, and the differentiating question is why. Are those sources following a format, answering comprehensively, or offering contrarian takes others do not? Digging into that common thread, Matt contends, is what separates successful marketers from the rest.
Where the Game Is Actually Won
The data points to a counterintuitive hierarchy of content types. Surfer analyzed 289, 15 URLs cited across ChatGPT, AI Overviews, AI mode, and Perplexity and found blog posts make up only about 29% of cited sources, yet mentions in blogs, both your own and third-party, correlate most strongly with how highly AI recommends a brand, more than reviews, news articles, or any other content type. Even though blogs are a minority of what gets cited, Matt concludes they are disproportionately where the game is won.
For teams who want to skip the copy-pasting, tools like Surfer's AI tracker automate the loop: monitoring brands across ChatGPT, Gemini, and Perplexity daily, producing visibility scores, mention rates, and a sources dashboard listing every domain the models pull from. Matt emphasizes that Surfer scrapes real web interfaces rather than APIs, because otherwise you may be tracking data your customers never see. In his demonstration with InVideo, the mention-gap view (places where AI cites rival Synthesia Studio but not InVideo) becomes, in his words, a blueprint for AI visibility on a silver platter.
Getting Into the Sources AI Trusts
Once the target sources are known, the final phase is earning a place in them. Matt lays out seven strategies in rough order of impact, beginning with the biggest lever: getting featured on third-party authority sites. ChatGPT's most-cited sources are Reddit, Wikipedia, and a cluster of well-known publications and review platforms, and much of what drives AI mentions is what other trusted sites say about you, not your own website. Since a competitor will not list a rival, he suggests pitching comparison publishers directly, perhaps offering demos or free access; even without a backlink, the brand mention itself has value because large language models read the page and sample its information.
The second strategy flips control: create content that becomes the source. If one particular source keeps surfacing across prompts, that format and angle work, so cover it yourself, ideally more comprehensively, with more current data. Roundup and comparison posts are singled out as formats AI loves to cite, and Matt dismisses the bias objection: as long as the piece is honest about pros and cons and genuinely helpful, both Google and AI models will reward it. Readers expect a comparison on a company blog to be useful, not perfectly neutral.
Reddit deserves special care. In a TIZenith study of about 187 technical queries, ChatGPT cited Reddit in 81% of its answers (but the cited posts were not the most upvoted ones. AI tends to ignore viral posts and prefer definitive, specific answers, and the median age of a cited post is over a year and a half. The implication is evergreen helpfulness, not trend-chasing: engage in the community for weeks without mentioning your brand, and if a natural mention is impossible, stay out) being shamed for astroturfing is worse than absence.
Structure, Clusters, and Being Everywhere
The remaining strategies compound each other. Wikipedia accounts for roughly 8% of ChatGPT citations, climbing as high as 14% during parts of 2025, but its notability bar is strict, even without a full article, a business can add itself to Wikidata, the structured database powering knowledge panels, which is a lower bar that still helps AI understand what a brand is. On-site, content should be built for extraction: clear definitions up front instead of marketing fluff, semantic HTML with tables for comparisons and pricing, and FAQ sections with concise, direct answers. As Matt admits, this is mostly just good SEO.
Topical depth matters more than single pages. Surfer's study of AI Overviews found they mention an average of five sources per query, and 90% of the time list eight or fewer, but those sources can come from multiple pieces of content on the same domain. One article gets one shot; a full cluster of pillar pages, supporting articles, and comparison guides can appear several times in a single AI response. In his dashboard demo, one domain showed eight commonly cited URLs across the prompts he tracked.
The final principle is ubiquity: a presence on Google Business Profile, LinkedIn, Crunchbase, and industry directories, with consistent brand information across all of them. The more widespread and consistent the descriptions, the more confident a model can be about mentioning a brand, a dozen consistent mentions beat one or two. The overall playbook, Matt sums up, is that competitors showing up without you is useful information: they are showing you the sources, content types, and platforms the AI trusts, and all of it is visible if you know where to look.