Writesonic went from 2.5% of leads via AI search to 35% in under a year. Samanyou Garg, the company's founder and CEO, laid out the system behind that number in a recent Search Engine Journal webinar. The uncomfortable part: 96% of AI search citations pointed to third-party sources. Reddit, YouTube, forums, industry publications. Not your website.
A few months earlier that figure sat near 80%. It's accelerating in the wrong direction for teams still treating their .com as the center of the SEO universe.
The CTR Collapse Is Real, but the Opportunity Isn't Zero-Sum
AI Overviews now appear in roughly 50% of U.S. Google searches, up from 6.49% in January 2025. When an AI Overview is present, users click traditional results about 8% of the time. Without one, that number is 15%. Organic CTR on queries with AI Overviews dropped from 1.76% to 0.61%.
Those numbers look grim in isolation. But there's a counter-signal worth watching: total search usage (engines plus LLMs) increased 26% worldwide. Google Search sessions actually rose to 12.6 per week among ChatGPT adopters, up from 10.5 before. Users refine ideas in chatbots, then return to Google with sharper queries. The pie got bigger. The slices just got rearranged.
Traditional SEO traffic share is projected to grow from 45% in 2025 to 53% in 2026 as that zig-zag behavior matures. So this isn't a funeral for organic. It's a restructuring.
Three Signals AI Agents Actually Care About
Search is shifting from keyword matching to task completion. AI agents evaluate, synthesize, and act on behalf of users. The signals that matter now aren't the ones most teams are optimizing for.
Experts identify three: entity clarity (who you are, what you do, where you operate), topical authority (depth and consistency across a subject), and trust and engagement (reviews, time on page, return visits, authoritative backlinks). If an AI agent can't parse your entity cleanly, your content quality is irrelevant. Slow, broken, or poorly structured pages won't get surfaced regardless of how good the writing is.
Structured data at scale (schema for services, products, reviews, FAQs, locations) is the immediate operational fix. Not glamorous. Effective.
The Multi-Platform Problem Nobody's Staffed For
Brand recommendations overlap less than 15% across AI platforms. What ChatGPT recommends, Perplexity and Google's AI mostly don't. "You need to make sure you are not putting all of your eggs in one basket, like your own website or a specific website," Garg said in the webinar.
This has real GTM implications. If your SEO team is measuring rankings on Google alone, they're missing the majority of AI-driven discovery. AI referral traffic grew 527% year-over-year. Gen AI traffic is growing 165x faster than organic search traffic. ChatGPT alone has 900 million weekly users and accounts for about 20% of global search-related traffic.
The measurement gap is the dangerous part. Google launched native AI search visibility reporting in Merchant Center (Share of Voice, Funnel Performance), which signals AI visibility as a distinct KPI category. But most teams haven't wired this into their dashboards yet. If you can't measure AI citations and brand mentions alongside traditional rankings, you're flying partially blind on a channel that's growing faster than anything else in the stack.
Closed-Loop SEO: Treat Every Page as an Experiment
Garg's team runs what he calls closed-loop SEO. Every published page gets a hypothesis. Confirm Google indexed it. Confirm it ranks or earns citations. Feed the result into the next iteration. They score pages with a business impact potential formula using four weighted factors that turn a 100-page backlog into a ranked work queue.
"Diagnosis is cheap now," Garg said. "The main thing is execution."
The webinar's live poll showed most attendees measure nothing, or measure and act on none of it. That gap between diagnosis and execution is where pipeline leaks. Google's core updates are already targeting "scaled content abuse" and low-value pages. Generic programmatic content is getting riskier, especially in SaaS categories with many similar pages. Authority-building content systems that AI agents can trust and cite are the defensible play.
What to Do This Week
Setup: Audit your top 20 commercial queries across Google AI Overviews, ChatGPT, and Perplexity. Note where competitors appear and you don't. Garg's team built an agent for this; you can start with manual spot-checks.
Hypothesis: If we add structured schema (product, FAQ, review) to our top 10 landing pages, then AI citation frequency for our brand will increase within 60 days, because entity clarity is a primary signal for agent-driven search.
Success metric: AI citations or brand mentions for target queries (tracked via Google Merchant Center AI reporting plus manual monitoring). Guardrail: Organic traffic to those pages doesn't drop more than 10%. Stop-loss: If organic traffic drops 15%+ within 30 days, revert changes and diagnose.
The trade-off: This front-loads ops work on schema and technical hygiene. It won't produce a visible lift in traditional ranking metrics immediately. The leading indicator is whether AI platforms start referencing your brand in responses to target prompts.
Garg's Writesonic data showed that if a brand is cited in an AI Overview, its organic CTR runs 35% higher than non-cited brands. The old game was ranking. The new game is being the source AI agents trust enough to cite. Same content, different engineering problem. And the teams that treat it as an engineering problem first are the ones capturing pipeline from a channel that barely existed 18 months ago.
