In mid-July 2026, an antitrust filing put a number on the table: Google's AI Overviews cut click-through on the top-ranked page by about 58%. Not a forecast — a traffic collapse that has already happened. Paired with Ahrefs' February 2026 study — AI Overviews now appear on roughly 48% of results, up 58% year over year — the old Google-search-sends-traffic model is being eaten by a layer of AI summary.
The shift over 12 to 18 months is structural. Google first wedged AI summaries above results, then launched Gemini-powered AI Mode, replacing the ten blue links with conversation. How real is the damage? Per public data, tech publisher Digital Trends fell from 8.5M monthly clicks in March 2024 to about 265K in January 2026 — 97% evaporated; The Verge, HowToGeek and ZDNet each lost over 85% in the same window. Not seasonal noise — a behavior change where users read the summary and never click. Market-wide, AI-summary coverage is past 40% and still growing.
Meanwhile, traffic is dispersing to other entry points. ChatGPT holds about 87% of AI referral traffic, Perplexity 12–15%. Citation behavior differs sharply: Perplexity attaches sources about 97% of the time, ChatGPT only about 16%. So even within AI traffic, whether you get clicked back varies wildly by platform. The category now reads: Google uses AI to retain users, other AI entry points split the rest. Search is not dead — it fragmented.
For a Taiwanese SME owner or dev studio, this is close: if your site's organic traffic quietly declined these two years, you probably didn't break your SEO — the SERP real estate got taken by AI summaries. Here is what to do and which spending to skip.
Event Details + Full Numbers
Facing a wave of criticism and lawsuits, Google in 2026 shipped several send-traffic-back updates, including a Further Exploration curated-links section inside AI Mode/AI Overviews and labels for subscription content. The official line is returning clicks to sites it is accused of cannibalizing. But the core numbers didn't improve: AI summaries still appear on nearly half of queries, top-page CTR still sinks about 58%.
Worse is the you-can't-even-see-it problem. Performance and experience are measurable, but AI traffic reflects less than 1% of real AI activity in default GA4 reports — many AI-assistant citations lack a referrer and get bucketed as direct or vanish. That means two things: your traffic drop may be understated, and your AI-citation success may be badly understated too. Fix measurement before optimizing.
Immediate Actions for Three Reader Types
- Brand owners / SME owners: Stop grading marketing only by Google rank. Open ChatGPT, Perplexity and Google AI Mode, ask 5–10 questions a customer would about your brand and category, and log whether you're mentioned and linked. That list is your new KPI baseline.
- Marketers / SEOs: Restructure content to be directly citable — clear Q&A headings, self-contained paragraphs, explicit numbers and definitions, and structured data (FAQPage, Article). AI summaries prefer content that lifts cleanly into an answer.
- Developers / studios: Add JSON-LD, server-side rendering (SSR) and clean semantic HTML for clients, and set GA4 custom channels to manually segment AI-assistant referrers so how-much-AI-traffic becomes measurable.
SaaS Tool Comparison
| Tool | Positioning / Pros | Cons | Price tier (approx) |
|---|---|---|---|
| Profound | Enterprise AI-visibility monitoring, multi-engine, deep data | Expensive, big-brand-leaning | High (enterprise) |
| Bluefish | AEO monitoring and competitor comparison | Overlapping features, learning curve | Mid-high |
| Scrunch AI | AI-search visibility tracking, clear reports | Still pricey for SMEs | Mid |
| Peec AI | Lightweight, good entry point for small teams | Limited engine coverage/depth | Low-mid |
What They Won't Tell You
- The send-traffic-back updates are mostly cosmetic: Further Exploration links and subscription labels don't change the read-the-summary-and-leave behavior. Waiting for Google's conscience is not a strategy.
- AI citation is highly unstable: The same brand may be cited now and gone next time. Betting the whole budget on one AEO SaaS dashboard means paying a high monthly fee for a jittery metric.
- Cited is not sold: AI traffic does convert well (ChatGPT ~14–16%, far above Google organic ~1.76%), but only if you get clicked back — ChatGPT attaches links to only ~16% of answers, so much exposure never returns.
The No-SaaS-Subscription SMB Alternative
You don't need enterprise monitoring first. Use a spreadsheet plus a schedule to run low-cost AI-visibility tracking yourself: list 10–20 questions a customer would ask, query ChatGPT, Perplexity and Google AI Mode once a week, and log mentioned / linked / position. To automate, use an open-source script to call each API periodically and write results to a database for a small in-house dashboard. This DIY version costs almost nothing per month yet answers the key question: over the last two months, is AI mentioning my brand more.
FAQ
My organic traffic dropped — did I break my SEO?
Not necessarily. In 2026 much traffic loss comes from AI summaries taking SERP space and users not clicking. Confirm your visibility inside AI summaries and AI assistants first, then decide whether it's a content problem or a real-estate problem — don't cut content that was working.
So should I still do SEO?
Yes, but broaden the goal. Traditional SEO gets you into rankings; AEO/GEO gets you into AI's answers. Their technical base overlaps heavily: good structured data, clean semantic HTML and citable paragraphs help both.
How do I know if AI actually brings me customers?
Build GA4 custom channels to segment ChatGPT/Perplexity referrers (default reports miss most AI traffic), then cross-reference with the weekly manual-query log to link cited with converted.
My Take
The mainstream advice is buy an AEO tool and chase AI visibility. My read is more reserved: for most Taiwanese SMEs in 2026, the best value isn't buying a dashboard — it's raising content and structured data to a worth-citing level and tracking metrics yourself at zero cost. AEO SaaS monthly fees rarely pencil out for a small site with dozens of organic visits a day, and AI citation is so jittery that paying for a bouncing score invites anxiety spending. The implication for ScriptWalker is direct: this is a tech-plus-content-structure service opportunity — getting SSR, JSON-LD, FAQ structure and GA4 segmentation right in one pass is more durable, and more budget-appropriate, than reselling a subscription dashboard. Search fragmented, but making content machine-readable became more valuable, not less.
Sources
- Google official Search blog (first-hand): https://blog.google/products/search/
- Google Search Central developer docs (first-hand): https://developers.google.com/search
- web.dev official performance/experience docs (first-hand): https://web.dev/articles/lcp
- Search Engine Journal: AI Overviews impact on publishers and the Ahrefs study (third-party): https://www.searchenginejournal.com/impact-of-ai-overviews-how-publishers-need-to-adapt/556843/
- The Next Web: Google updates AI Overviews amid 58% click decline (third-party): https://thenextweb.com/news/google-ai-overviews-publisher-links-search-traffic