AI search strategy no longer revolves around page-one rankings. A Semrush study published in 2025 found that nearly 90% of the pages ChatGPT cites in responses rank at position 21 or lower in traditional organic search for related queries. The front door of search is being rebuilt, and the old ranking leaderboard barely applies to the new system.
Why Does a Page-One Google Ranking Not Guarantee an AI Citation?
When a user asks ChatGPT, Perplexity, Gemini, or Google AI Mode (Google’s conversational search experience, which fully replaces the traditional results page) a buying question, the assistant reads dozens of sources, composes a single answer, and names a small number of brands. The link to your page is optional. Being the most-visited page in your category does not make you the most-cited.
A practitioner who spent early career years at Google scaling the Ads platform, and now runs a growth consultancy with AI search visibility programs for more than a dozen VC- and PE-backed startups, documented this directly in a Search Engine Journal piece. One client held page-one positions across its category and did not appear in a single AI-generated answer for its core buying queries. A second client had middling Google rankings but a strong footprint on Reddit and in trade publications, and was cited by AI assistants “constantly, far out of proportion to their rankings.”
Semrush’s 2025 research confirms the pattern at scale: citation authority and ranking authority are not the same asset.
What Six Signals Actually Drive AI Citations?
Based on audits described in the Search Engine Journal piece, six factors explain most of what determines whether an AI assistant names a brand in its response:
- Brand Authority: Assistants favor brands that third parties mention by name on trusted sources. Content published on your own website carries far less weight than unprompted external mentions.
- Content Freshness: Pages untouched for two or more years are frequently skipped. Visible, dated updates matter more here than they typically did for classical SEO.
- Entity Recognition: AI systems process knowledge in terms of entities (named people, companies, and concepts with established identities across the web). Brands with recognized expert voices and consistent descriptions across multiple platforms are easier for assistants to reference confidently.
- Extraction-Ready Structure: A direct answer near the top of the page, clear headings, and schema markup (structured metadata added to a page’s code so machines understand what it is about) help AI systems pull specific claims accurately. Google’s Danny Sullivan has noted that structured data “supports understanding rather than guaranteeing a win,” which aligns with what the SEJ audits show.
- Cross-Platform Consensus: When Reddit threads, YouTube videos, review platforms, and press coverage describe a brand in consistent terms, AI systems read that agreement as a trust signal. This is the hardest signal for a competitor to replicate quickly.
- Earned Media: Coverage the brand did not commission or write itself carries the most weight in what gets cited.
As the SEJ piece notes, this list is mostly brand and PR work applied with performance marketing discipline. An AI search strategy that lives entirely on your own website is working the smallest part of the system.
How Do You Start Measuring AI Search Visibility?
The SEJ piece offers a concrete starting audit. Write five prompts a real buyer would use, such as “best [category] for [use case]” or “[your product] vs. [competitor].” Run each across ChatGPT, Perplexity, Gemini, and Google AI Mode: 20 runs total. Log each result as named, cited with a link, or absent. Appearing in fewer than 12 of the 20 indicates a measurable citation gap. For ongoing monitoring, tools including Profound, Otterly.AI, and Peec AI track citations across platforms over time, turning a one-time snapshot into a trendline.
Semrush’s 2025 research adds urgency to that measurement. AI search visitors currently convert at 4.4 times the rate of the average traditional organic search visitor, based on Semrush’s conversion-rate analysis. The research projects that AI search traffic could match traditional organic search in total economic value globally by the end of 2027. ChatGPT’s weekly active users grew 8x between October 2023 and April 2025, reaching over 800 million weekly active users. Semrush also projects AI search could surpass traditional search in total visitors to websites by early 2028, with that timeline potentially compressed if Google’s default experience shifts fully to AI Mode.
What This Means for AI-Search Visibility
Here is the simplest way to read the data: ranking well in Google and being trusted by AI assistants are two separate reputations, and most brands have only been building one of them.
Picture a buyer asking an AI assistant: “What’s the best logistics software for a 50-person warehouse?” The assistant does not check your Google ranking. It looks for brands that appear repeatedly across forums, trade publications, and review sites, described in consistent terms, mentioned by sources it already has reason to trust. A brand that holds page-one positions but exists mostly on its own site is, from the assistant’s perspective, a stranger to the community.
The Semrush finding that nearly 90% of ChatGPT-cited pages sit at position 21 or lower sounds like a failure in traditional SEO terms. In AI citation terms, it means those pages succeeded in earning mentions and links from external sources, which is exactly what AI systems weight most. The ranking ladder and the citation network overlap partly, but they have different winners at the top.
One implication neither source spells out explicitly: standard marketing dashboards tracking organic traffic and conversions have no visibility into the citation gap. A brand can be losing buying-stage consideration to AI-recommended competitors without that loss appearing anywhere in its Google Search Console data. That is a monitoring blind spot, not just a strategy gap. Brands often discover it only when they run the 20-prompt audit and find a competitor named in 18 of 20 responses while they appear in four.
Hingewise’s assessment: brands that have invested heavily in owned-channel content but underinvested in earned media, community presence, and third-party coverage are more exposed to AI search displacement than their traditional rankings suggest. The 4.4x conversion premium Semrush identifies also means the exposure is not symmetric: losing AI citation share to a competitor costs more per lost buyer than losing an equivalent number of clicks in traditional search.
Five Checks Before Assuming Your AI Visibility Is Fine
- Run the 20-prompt drill: five buyer questions across ChatGPT, Perplexity, Gemini, and Google AI Mode. Log each as named, cited with a link, or absent.
- Check whether your brand appears on the third-party sources your buyers actually use: Reddit communities, G2 (or whichever review platform your category trusts), trade publications, and industry forums.
- Review your top revenue pages: do they open with a direct answer to the buyer’s question, or is the key claim buried several paragraphs in?
- Audit how consistently your brand is described across platforms. Conflicting descriptions give AI systems inconsistent entity signals and reduce citation confidence.
- Check when your highest-traffic pages were last substantively updated, and whether a visible revision date or update note appears on the page.
What to Watch Next
The SEJ piece frames AI search as the latest in a line of platform shifts: broad match, mobile, automated bidding. Each rewarded the teams that moved earliest. What current data does not yet answer is whether AI citation authority, once earned, compounds the way domain authority did in traditional SEO, or whether the landscape stays more fluid as AI systems update their training and retrieval (the process by which an AI pulls relevant content when answering a query) more frequently than Google updated its index. As tracking tools mature and citation data becomes comparable across platforms, that question should become answerable within the next 12 to 18 months.
Lam Nguyen · Hingewise
Sources:
Search Engine Journal
Semrush AI Search Traffic Study, 2025