Why this matters now, not eventually
Three data points make the case better than any prediction could:
- AI search engines are cutting into traditional clicks. Ahrefs analyzed 300,000 keywords and found that when a Google AI Overview appears above the search results, the top-ranking organic page’s click-through rate drops by roughly 34.5% compared to similar searches without an AI Overview. People are getting their answer without ever clicking through.
- Local discovery is shifting to AI tools too. BrightLocal’s 2026 Local Consumer Review Survey found that the share of consumers using AI tools like ChatGPT and Google AI Mode to discover local businesses jumped from 6% in 2025 to 45% in 2026, making AI the third most common discovery channel, ahead of Yelp and Tripadvisor, behind only Google and Facebook.
- Being cited is a learnable skill, not luck. Researchers from Princeton, Georgia Tech, the Allen Institute for AI, and IIT Delhi published the first controlled study on this, introducing what they call Generative Engine Optimization. Using a 10,000-query benchmark across nine domains, they found that applying specific content tactics could boost a page’s visibility in AI-generated answers by up to 40%.
Put together: fewer people are clicking through from Google for informational searches, more people are asking AI tools directly, and there’s now peer-reviewed evidence that how you structure your content changes whether AI tools cite you. That’s worth paying attention to.
What the research says actually gets cited
The most detailed study on this came from AirOps, which analyzed over 12,000 URLs, comparing pages ChatGPT cited in response to 900 real queries against the pages ranking on page one of Google for those same queries. The gap between the two groups was consistent and specific, not vague:
- Structured data (schema markup). 61% of ChatGPT-cited pages used rich schema markup, versus just 25% of Google’s top-ranking pages. FAQ schema specifically was more than twice as common on cited pages (10.5% vs 5.4%).
- Clean heading structure. 68.7% of ChatGPT-cited pages followed a proper sequential heading order (H1, then H2s, then H3s, with nothing skipped or out of order), compared to only 23.9% of Google’s top-ranking pages. A single, clear H1 was present on 87% of cited pages versus 64% of Google’s top results.
- Lists. Roughly 80% of ChatGPT-cited pages included at least one section formatted as a list, compared to 28.6% of Google’s top-ranking pages. Cited pages also averaged far more list sections per page.
The pattern across all three findings is the same: AI tools favour content that’s easy to pull apart into discrete, extractable pieces. A wall of unbroken paragraphs might read fine to a person, but it gives an AI system very little to grab onto.
What Google itself says about this
It’s worth being precise here, because a lot of AI-search advice overstates what’s actually confirmed. Google’s own Search Central documentation is direct on this point: there is no special schema.org markup required to appear in AI Overviews or AI Mode. AI Overviews draw from the same underlying Search index as regular results. Google’s guidance is to keep doing standard structured data correctly, and to focus on making content that’s genuinely useful rather than written to game a system. Their own guide to optimizing for generative AI features repeats the same core message as the third-party research above: clear structure and genuinely helpful content are what carry over into AI-generated answers, not tricks.
Practical steps for a small business website
None of this requires a rebuild. It’s mostly about how existing content is structured:
- Use one clear H1 per page, and keep headings in proper order underneath it (H2s, then H3s), rather than skipping levels or using headings for visual styling.
- Answer the question first, then explain. Whatever question a page or section heading implies, answer it directly in the first sentence or two that follows. Save the elaboration and nuance for after.
- Add FAQ sections with FAQ schema on service pages and blog posts where real questions come up. This was one of the clearer differentiators in the AirOps data.
- Use real lists (proper HTML bullet points or numbered lists) for steps, features, or comparisons, instead of describing the same information as a paragraph.
- Keep your Google Business Profile complete and current if you’re a local business. Given how much local discovery is shifting toward AI tools, the same profile completeness that helps in Google’s local pack is part of what AI tools draw on for local recommendations too.
- Don’t abandon standard SEO. Google’s own guidance is explicit that this isn’t a separate system: the same structured data and helpful-content practices that support regular search rankings are what carry over into AI-generated answers.
The bottom line
AI search isn’t replacing Google search, but it’s clearly becoming a second, parallel path customers use to find and vet a business, and the two paths don’t always reward the same pages. A page can rank well on Google and still be invisible inside an AI-generated answer if it’s a dense wall of text with no schema, no clear headings, and no lists for an AI system to extract from. The fix isn’t exotic: clear structure, direct answers, real schema markup, and content built to be genuinely useful, the same fundamentals good SEO has always rewarded, just applied with AI systems in mind as a second reader alongside Google and your actual visitors.
