Zero-Click Search: How to Prove Marketing Value Now

Your website traffic can drop while your business keeps growing, and that’s not bad luck. It’s zero-click search doing exactly what it’s built to do. In the first four months of 2026, 68.01% of U.S. Google searches ended without a single click to a website, up from 60.45% in 2024 ^[1]. That figure doesn’t even count searches made inside Google’s mobile app, where SparkToro expects the zero-click rate is higher still. That’s not a dip. It’s a structural shift in where the customer journey actually happens, and it means the old scoreboard (sessions, pageviews, bounce rate) no longer tells you whether your marketing is working.

For small businesses and nonprofits watching every dollar, this is scary territory. You can’t afford to guess. But you also can’t keep reporting on a metric that AI search has quietly hollowed out. The fix isn’t abandoning measurement. It’s measuring the right things, consistently, and being honest about what the numbers can and can’t prove.

Why Your Traffic Dashboard Is Lying to You

Pew Research Center tracked the real browsing behavior of 900 U.S. adults across 68,879 Google searches. When an AI summary appeared at the top of the page, people clicked a regular search result in just 8% of visits. Without the summary, that nearly doubled to 15%. Only 1% of visits included a click on a source link inside the AI summary itself, and people were more likely to end their browsing session entirely after seeing one (26% of the time versus 16%) ^[2]. The answer satisfies them right there. They never leave the search page.

That was observational data. In March 2026, researchers from the University of Pennsylvania and Northeastern University ran a preregistered field experiment with 1,100 U.S. Google users to test cause and effect. Forcing every search into Google’s AI Mode cut clicks to outside websites by 18.8 percentage points, and those users turned to Bing, DuckDuckGo, or Yahoo 11.2 percentage points more often ^[3]. Hiding AI Overviews raised click-through by an estimated 8.8 percentage points. One honest caveat: Google changed its page code mid-study, so the researchers’ blocker only caught about half of the Overviews, which means that 8.8-point figure is an adjusted estimate rather than a clean measurement ^[4]. People aren’t leaving the funnel. They’re finishing more of it before they ever reach you.

Position one isn’t safe either. Ahrefs compared 300,000 keywords between December 2023 and December 2025. For keywords that now trigger an AI Overview, the top-ranking page’s click-through rate fell to about 0.016, against a predicted 0.037 if Overviews had never launched. That’s roughly 58% fewer clicks, up from the 34.5% loss Ahrefs measured just eight months earlier ^[5]. Put plainly: for every 100 clicks a #1 ranking used to earn, Google now keeps 58.

Getting cited inside the AI answer is the closest thing to a lifeline. Seer Interactive tracked 53 brands across 5.47 million queries and found that brands cited inside an AI Overview earned roughly 120% more organic clicks per impression than brands on the same results page that weren’t cited ^[6]. Seer is careful to call that a correlation, not proof of cause, since stronger brands are also more likely to get cited in the first place. And ranking well doesn’t guarantee the citation: in Ahrefs’ latest analysis of 863,000 keywords, only about 38% of AI Overview citations came from pages ranking in Google’s top 10 for that search ^[7]. Ranking and getting cited are now two different jobs.

The Scoreboard You Actually Need

If traffic alone can’t tell the story, what can? Four things, tracked together:

• Qualified leads and conversions, not raw visits. Form fills, calls, bookings, donations, and registrations. Set them up as key events in GA4 and tie them to your CRM or donor system so you can see revenue, not just activity.
• Branded search volume. The people typing your business name directly. In Google Search Console, filter Performance by queries containing your name (and common misspellings) and watch impressions and clicks month over month.
• Assisted influence. AI exposure that shows up later as a conversion. You’ll catch it with a “How did you hear about us?” question and by segmenting referrals from chatgpt.com, perplexity.ai, gemini.google.com, and copilot.microsoft.com in your analytics.
• Brand mentions and recommendations across a fixed set of customer questions you track over time (more on building that below).

Here’s why this matters so much for local and nonprofit organizations specifically. The first large zero-click study built on South African data, from performance agency Algorithm, analyzed 16 months of Search Console data from 110 websites run by 71 businesses. The typical business went from 4.8 to 3.0 clicks per 100 appearances in Google, a 29% drop, even as it showed up in mobile results 22% more often. Branded searches were the bright spot: they converted at 15.6 clicks per 100 appearances and made up only about 13% of appearances but roughly 75% of clicks ^[8]. Visibility and traffic have split apart, and your own name is where the clicks still live. If you run a youth sports league or a regional service business, this is already happening to you, whether your reports show it or not.

The honest answer is that no single number replaces the old traffic report. You need a system, not a stat.

The Invisible Assist: How AI Recommendations Show Up as “Branded Search”

Here’s a pattern that trips up a lot of reporting. Similarweb tracked real user journeys across finance, travel, and beauty and found that people who got a brand recommendation from ChatGPT were 2.5 times more likely to visit that brand’s website within seven days than a competitor’s site. But roughly 56% of those AI-influenced visits arrived through search, compared with about 40% of ordinary visits, meaning people typed the business name into Google rather than clicking a link inside the AI answer ^[9]. Your analytics tool will label that as organic or direct traffic. It won’t tell you the AI recommendation caused it.

Growth advisor Kevin Indig went looking for exactly this effect, and his results are a good reason to stay careful. Across GA4 data from 12 sites, Direct traffic looked like it was booming, but engaged Direct sessions were basically flat, which points to bots rather than hidden AI visitors. In a separate analysis of four brands, he ran 632 tests comparing weekly AI visibility against branded search and Direct traffic at delays of zero to six weeks, and he couldn’t establish a reliable link. One site did show a 0.72 correlation between AI share of voice and Direct traffic at a three-week lag, but the pattern didn’t hold across the group ^[10]. Others read that same data more optimistically, arguing that brand channels like billboards have never been cleanly attributable either, and that the strongest individual results suggest a real but hard-to-isolate effect ^[11].

So don’t oversell this. AI exposure looks like it drives delayed, branded demand in some cases, but it’s not proven causation across the board. Treat it as a signal worth watching, not a guarantee worth promising a client or a board.

What can you actually do about it? Add a simple “How did you hear about us?” field to every intake form, donation page, and booking flow, and include “ChatGPT or another AI assistant” as an option alongside Google, social media, and word of mouth. It’s low-tech, but it’s the most honest way to catch demand that your analytics tool will otherwise misfile. And protect your branded search results: make sure your Google Business Profile, homepage, and top service pages are complete and current, because that’s where AI-influenced demand lands.

Why a Single ChatGPT Screenshot Proves Nothing

A lot of marketing reports now include a screenshot: “Look, we showed up in ChatGPT!” It feels like proof. It isn’t.

Rand Fishkin of SparkToro and Patrick O’Donnell of Gumshoe.ai tested this directly. About 600 volunteers ran 12 prompts through ChatGPT, Claude, and Google’s AI Overviews (or AI Mode when no Overview appeared) for 2,961 total runs in late 2025. The odds of getting the same list of recommended brands twice were less than 1 in 100. The odds of getting the same list in the same order were less than 1 in 1,000 ^[12]. The encouraging part: when they tested 142 different human-written versions of a headphone-shopping question, the strongest brands still appeared in 55% to 77% of the 994 responses. How often you show up across many runs is a meaningful number. Your “rank” in any single answer is not.

Academic researchers have reached the same conclusion. A 2026 paper titled “Don’t Measure Once” tested repeated runs across ChatGPT, Perplexity, Gemini, and Google AI Mode and recommended at least 7 runs per prompt for brand visibility monitoring, and at least 8 when you care about which sources get cited ^[13]. Below that, you’re reporting noise and calling it a trend.

It gets more fragile still. Ahrefs compared 730,000 pairs of AI Mode and AI Overview answers for the same searches. The answers reached essentially the same conclusions (86% semantic similarity), yet cited the same URLs only 13.7% of the time ^[14]. And citations don’t stay put once you earn them: in a separate study of 43,000 keywords, Ahrefs found AI Overview content changes about every two days, and between one capture and the next, 45.5% of the cited sources were entirely new ^[15].

Wording and format matter too, though maybe not the way you’d expect. A controlled study by Peec AI found that the intent behind a question matters more than its exact words, but asking for a list or a ranking, or using short keyword-style prompts, surfaced up to 20% to 25% more brands than open-ended questions ^[16]. If your measurement strategy depends on one phrasing of one question, run once, you’re building on sand.

Building a Customer Question Panel That Actually Holds Up

So what does reliable measurement look like for a business that doesn’t have a data science team? Look at how the big players do it. Semrush built its 2026 AI Visibility Index around fixed monthly pools of 2.5 million U.K. prompts and 25 million U.S. prompts, run the same way each month across ChatGPT, Gemini, and Google AI Mode, with AI Overviews tracked separately ^[17]. You don’t need millions of prompts. You need the same discipline at a smaller scale.

1. Write a frozen list of 25 to 50 real customer questions. Pull them from sales calls, intake forms, reviews, and the questions your front desk answers every week. Avoid putting your own business name in them; a question that names you will almost always mention you, which tells you nothing.
2. Tag each question by service line, local intent (your city or region), competitor comparisons, and how close the person asking is to making a decision.
3. Run the list across 3 to 5 AI platforms (for most local businesses: ChatGPT, Google AI Overviews, Google AI Mode, Gemini, and Perplexity), using fresh, logged-out sessions from the same location.
4. Repeat each question at least 7 times per cycle, on a monthly or quarterly schedule. Don’t change the wording or format between cycles. That’s the whole point.
5. Log every run in a simple spreadsheet: date, platform, question ID, run number, whether you were mentioned, any URL cited, your position, which competitors appeared, and anything the AI got wrong about you.

Once you have that panel running, split what you’re tracking into separate metrics instead of lumping them together:

• Mention rate: does the AI answer name your brand at all?
• Citation rate: does the answer link to your site or to an earned third-party source about you?
• Recommendation position: are you the leading suggestion or an afterthought?
• Accuracy: are your hours, services, prices, and service area described correctly?

A brand can be mentioned without ever being recommended. It can be recommended without ever getting a click. Treating these as one blended “AI visibility score” hides exactly the information you need to act on. (We dug into why mentions and citations so rarely line up in Why AI Recommends Your Competitors, Not You.) When you report results, show the sample size: “mentioned in 18 of 35 runs” is more honest, and more useful, than “51% visibility.”

What This Means for Resource-Strapped Organizations

None of this is optional anymore, even for a nonprofit with a skeleton marketing team or a small business owner doing their own content. Bain’s consumer research found that about 80% of consumers rely on zero-click results in at least 40% of their searches, which Bain estimates cuts organic web traffic by 15% to 25%. Its recommendation is to shift from click-focused metrics toward search impressions and AI reach ^[18]. That’s a full redefinition of how marketing gets measured, not a temporary SEO wrinkle. Businesses that keep reporting last decade’s metrics will keep making decisions off a map that no longer matches the territory.

The honest challenge is capacity. Running a 25-to-50-question panel across five AI platforms, at least seven times each, every month, logging results, and separating mention from citation from recommendation is real work. That’s somewhere between roughly 500 and 1,750 AI answers per cycle. Most small teams can’t do that by hand on top of everything else on their plate. That’s precisely the gap automated, AI-native systems are built to close, and it’s how we run oneninety8’s marketing engines: the question panel runs on schedule, the branded-search assist gets tracked, and all of it ties back to actual leads and conversions instead of vanity traffic.

Frequently Asked Questions

What is zero-click search?
Zero-click search happens when someone gets their answer directly on the search results page, often from an AI Overview, a featured snippet, or a map listing, without clicking through to any website. In the first four months of 2026, it was the outcome of 68.01% of U.S. Google searches ^[1].

How can I prove marketing ROI if people aren’t visiting my site?
Track qualified leads and conversions, branded search volume in Google Search Console, assisted influence from AI exposure (using a “How did you hear about us?” field), and your brand’s mention and citation rate across a fixed panel of customer questions. Traffic alone no longer tells the full story.

Is one mention in ChatGPT a sign my AI visibility is good?
No. A single screenshot or one-time appearance isn’t reliable. In SparkToro’s testing, the same prompt returned the same list of brands less than once in 100 runs ^[12]. You need repeated runs over time to see a real pattern, not a one-off result.

How many times should I run each AI prompt to measure visibility?
Research published in 2026 recommends at least 7 runs per prompt for brand visibility monitoring, and at least 8 if you’re tracking which sources get cited ^[13]. Keep the wording and format identical from one measurement cycle to the next.

Does showing up in an AI Overview actually drive traffic?
It can, but the relationship isn’t fully proven. Brands cited inside AI Overviews earned roughly 120% more organic clicks per impression than uncited brands in Seer Interactive’s data, though Seer notes that’s a correlation, not proven cause ^[6]. Broader testing has found no consistent link between AI visibility and later branded search or direct traffic ^[10]. Treat it as one signal among several, not a guarantee.

Does ranking #1 on Google still matter if AI Overviews take the clicks?
Yes, but it’s no longer enough. Top-ranking pages lose about 58% of their expected clicks when an AI Overview appears ^[5], and only about 38% of AI Overview citations come from pages ranking in the top 10 ^[7]. You need to rank, get cited, and protect your branded search results.

Your Next Step

Zero-click search isn’t going away, and pretending your old traffic reports still tell the full story won’t change that. The businesses and nonprofits that adapt fastest are the ones building a real measurement system now, one that tracks mentions, citations, branded search, and actual conversions together. Start small: pick 25 real customer questions, run each one at least seven times across a few AI platforms this month, and log what comes back. Add a “How did you hear about us?” field to your forms the same day. Those two moves will tell you more about where you stand than another quarter of watching pageviews slide.

Want a faster read? Our free Engine Teardown includes an AI visibility check: we run your customers’ real questions through ChatGPT and Google’s AI and show you who gets recommended, who gets cited, and where you’re missing. Or take the Marketing Engine Scorecard to see how your current setup measures up in a few minutes.

Sources

1. In 2026, Less than One Third of Google Searches Still Send a Click (sparktoro.com)
2. Google users are less likely to click on links when an AI summary appears in the results (pewresearch.org)
3. AI in Search Reduces Publisher Referrals Without Improving User Experience: Experimental Evidence (arxiv.org)
4. Google AI Mode cuts clicks and user satisfaction: Study (searchengineland.com)
5. Update: AI Overviews Reduce Clicks by 58% (ahrefs.com)
6. AIO Impact on Google CTR: 2026 Update (seerinteractive.com)
7. Update: 38% of AI Overview Citations Pull From The Top 10 (ahrefs.com)
8. South African businesses are being seen more in Google and clicked less: First local zero-click study (bizcommunity.com)
9. Being in AI Answers Drives 2.5x More Site Visits (similarweb.com)
10. Does AI drive more direct traffic? (growth-memo.com)
11. AI search is a branding channel, not a distraction (searchengineland.com)
12. NEW Research: AIs are highly inconsistent when recommending brands or products (sparktoro.com)
13. Don’t Measure Once: Measuring Visibility in AI Search (GEO) (arxiv.org)
14. Are AI Mode and AI Overviews Just Different Versions of the Same Answer? (730K Responses Studied) (ahrefs.com)
15. AI Overviews Change Every 2 Days (But Never Change Their Mind) (ahrefs.com)
16. What Matters In An AI Prompt? Intent or Keywords? (searchenginejournal.com)
17. Semrush 2026 AI Visibility Index Reveals Major UK-US Search Differences (martechedge.com)
18. Goodbye Clicks, Hello AI: Zero-Click Search Redefines Marketing (bain.com)

Researched and fact-checked against 18 sources, most of them original studies. Statistics verified October 30, 2026.