Are Websites Dead? What Consumers Say vs. What the Data Shows

Abstract

A 2026 consumer survey reports that more than half of people sometimes skip websites when researching with AI, and over 60% expect AI to eventually replace websites. This evidence review from Make It Loud sets that survey against observed browsing, click, traffic, and local engagement data to separate what consumers say from what they actually do, and explains what it means for a North Atlanta small business deciding how to judge and invest in its website.

Methodology

This report was prepared by Make It Loud Digital Marketing as part of its North Atlanta Small Business Research Series. It is an evidence review, not original research. It sets a 2026 Kinsta / Propeller Insights consumer survey (1,008 U.S. consumers, plus or minus 3 points) against observed behavior from Pew Research Center's 2025 browsing study (900 U.S. adults, 68,879 searches in March 2025), a randomized field experiment by Saharsh Agarwal and Ananya Sen, Reuters Institute / Chartbeat referral data (2,576 publisher sites), Sensor Tower's State of Web 2026, Birdeye's State of Google Business Profile 2026 (large multi-location brands), and BrightLocal's Consumer Search Behavior 2026 report (1,227 U.S. consumers). The Kinsta figures come from the company's published summary, which does not provide question wording, bases, or cross-tabulations, and Kinsta sells website hosting. The Agarwal and Sen experiment, Ahrefs, Reuters Institute / Chartbeat, and Sensor Tower findings were read through a trade-group summary from Digital Content Next and are labeled 'as reported.' Birdeye's findings are vendor research on large multi-location brands, and the causal link between AI answers and falling impressions is an interpretation, not a demonstrated result. The strongest click evidence describes publisher sites and general search, not local service businesses. No source reviewed measures the effect of AI exposure on the website traffic of a single-location small business, and no Georgia-specific data was found. This report does not constitute a guarantee of results for any individual business.

Findings

Surveys and behavioral data answer 'are websites dead?' differently because they measure different things. Kinsta's survey reports that 53.8% of consumers sometimes or often skip websites while researching with AI and that 61.2% expect AI to eventually replace websites, but these are self-reported habits and a prediction from a hosting company. Observed behavior partly supports the first claim: Pew found people clicked a traditional result in 8% of visits when an AI summary appeared versus 15% when it did not, and a randomized experiment found outbound organic clicks fell 39.8%. At the same time AI platforms send almost no traffic, with ChatGPT accounting for 0.02% of publisher referrals in Chartbeat data, so nothing yet replaces the clicks that are lost. Local businesses are a different case: in Birdeye's vendor data on large brands, impressions per location fell 53.8% while customer actions fell only about 5%, which commentators read as AI intercepting casual browsing while high-intent actions hold up. The practical lesson is to judge a website by calls, bookings, and inquiries rather than by pageviews.

Overview

“Are websites dead?” is a question that surveys and traffic data answer differently, because they measure different things. Surveys record what people say they do and believe. Browsing studies, experiments, and traffic logs record what people actually do. This report, prepared by Make It Loud as part of its North Atlanta Small Business Research Series, sets one against the other, using a 2026 Kinsta survey as the starting point and checking each of its headline claims against observed behavior.

A 2026 survey by hosting company Kinsta found that more than half of consumers sometimes skip websites when researching with AI, and over 60% expect AI to eventually replace websites as the main way people discover products. Observed behavior partly backs the first claim and says little about the second. Pew’s study of 900 adults’ actual browsing found people clicked a traditional result in 8% of visits when an AI summary appeared, versus 15% when it did not, and a randomized experiment found outbound organic clicks fell nearly 40%. At the same time, AI platforms still send almost no traffic. So websites are not dead, but the traffic that informational pages used to get from search is being intercepted, and nothing yet replaces it. For local businesses the picture is different: their websites are mainly where customers verify and act, and early data on Google Business Profiles shows impressions falling much faster than calls, directions, and website visits. The practical lesson is to stop judging a website by pageviews and start judging it by whether customers contact the business.

Finding 1: The survey claim, and what it does and does not measure

Kinsta’s survey of 1,008 U.S. consumers, conducted by Propeller Insights with a margin of error of plus or minus 3 points, reports that 59.2% use AI tools often or occasionally for research and that 53.8% sometimes or often skip websites entirely during that research. Another 61.2% believe AI will eventually replace websites as the main way people discover products. Among those who often rely on AI answers instead of visiting sites, 52.2% said AI is faster than searching multiple websites, 45.2% said it makes follow-up questions easier, 42% pointed to too many ads or pop-ups on websites, and 37.6% said traditional results feel cluttered or SEO-driven.

Several features limit how far these numbers can be pushed. The figures describe self-reported habits, not recorded behavior. “Sometimes” is a low bar, so a 53.8% figure that combines “sometimes” and “often” says little about how frequently sites are actually skipped, and the published summary does not state the base for that figure or break it down. The 61.2% figure is a prediction about the future, not a measurement. And the survey comes from a hosting company with an obvious interest in whether websites matter, so its framing deserves the same scrutiny as any vendor research. None of this makes the survey wrong. It makes it a claim to test, which is what the rest of this report does.

Finding 2: Observed behavior, clicks fall when AI answers appear

The most direct test of “people skip websites” is to watch what they actually click. Pew Research Center did that, analyzing the browsing activity of 900 U.S. adults who agreed to share it, covering 68,879 Google searches in March 2025. AI summaries appeared on about 18% of those searches, and 58% of participants saw at least one. Users clicked a traditional search result in 8% of visits when an AI summary appeared, compared with 15% when it did not. Clicks on links inside the AI summary itself occurred in just 1% of visits, and users ended their browsing session on 26% of pages with a summary, compared with 16% of pages without one. Google publicly disputed the study’s methodology.

Other evidence points the same way. A randomized field experiment by Saharsh Agarwal of the Indian School of Business and Ananya Sen of Carnegie Mellon University, covering more than 1,000 Google users, found outbound organic clicks fell 39.8% and zero-click searches rose 34.5% when an AI Overview appeared. Ahrefs, an SEO software vendor, analyzed 300,000 keywords and estimated an AI Overview reduced click-through to the top organic result by 58%. These studies use different methods, which is what makes the agreement meaningful, though the last two reached this report through a trade-group summary rather than the original papers.

Finding 3: AI sends almost no traffic, so nothing is replacing what is lost

If AI were replacing websites as the place people go, AI platforms would be sending visitors somewhere. The traffic data says they send very little. The Reuters Institute’s 2026 trends report, drawing on Chartbeat data from 2,576 publisher sites, found Google Search referrals fell 33% globally between November 2024 and November 2025, including a 38% decline in the U.S. Its authors caution that AI Overviews cannot account for all of that decline, since algorithm changes, publisher strategies, and shifts in audience behavior also play a role. In the same dataset, ChatGPT accounted for just 0.02% of publisher referrals and Perplexity 0.002%. Sensor Tower’s State of Web 2026 found traffic from AI assistants grew 86% year over year in 2025, yet generative AI referrals still made up only about 0.7% of overall web traffic.

The accurate reading is not that websites are dying but that something narrower is happening: AI answers satisfy some informational demand without a visit, and AI platforms are not yet sending enough visitors to offset it. These figures come from a summary published by Digital Content Next, a trade association of publishers, and describe publisher sites, so they should be read as strong evidence about informational content and weaker evidence about everything else.

Finding 4: Local businesses are a different case

Almost all the click evidence above concerns searches where an AI answer can fully satisfy the question, such as a news query or a how-to. A local service purchase usually ends in an action: a phone call, a directions request, a booking, a form. That action usually requires contacting the business, which is why local sites tend to work as verification and conversion pages rather than as destinations for answers.

Birdeye’s State of Google Business Profile 2026 report, based on large multi-location brands, found impressions per location fell 53.8% while customer actions fell only about 5%. Website visits made up 47% of profile engagement, direction requests 38%, and calls 15%. Birdeye and several commentators attribute the gap to AI answers intercepting casual browsing while leaving high-intent actions intact. That interpretation is plausible but not proven by the data, which is vendor research on large brands rather than small local businesses. Kinsta’s own survey adds a supporting detail: after receiving an AI recommendation, 44.7% always or most of the time visit the company’s website, and another 25% do so occasionally. No source reviewed for this report measures how AI exposure changes website traffic for a single-location North Atlanta service business, which is the evidence that would settle the question locally.

Finding 5: Reconciling what people say and what they do

The survey and the behavioral data are less contradictory than they first appear. Skipping a website while researching, which is what 53.8% of surveyed consumers report, is consistent with falling click rates on informational searches. Visiting a website to verify a company before contacting it, which 69.7% of surveyed consumers report doing at least occasionally after an AI recommendation, is consistent with stable customer actions on business profiles. These describe different stages of the same journey, and BrightLocal’s 2026 research on 1,227 U.S. consumers fits the same pattern: only 18% of consumers who used AI for local search felt ready to contact the recommended business, and 43% of those who started on an AI tool went on to check Google.

The sharper conclusion for a website owner is about measurement. If pageviews and Google impressions are the yardstick, a website will look like it is failing in 2026. If the yardstick is calls, direction requests, bookings, and form submissions, the evidence so far suggests the picture is considerably better. This is a hypothesis supported by suggestive evidence, not a settled finding, and it is worth testing against a business’s own analytics.

What this means for North Atlanta businesses

1. Judge your website by actions, not pageviews. Given that impressions and clicks are falling faster than customer actions in the available data, track calls, direction requests, form fills, and bookings alongside traffic, and judge changes by those outcomes.

2. Expect informational pages to be the most exposed. Content whose only job is to answer a general question is the content AI summaries most directly replace. A site’s value increasingly sits in the pages that help a customer verify and contact the business: services, service areas, proof of work, and clear contact paths. This is the kind of thinking behind our approach to website conversion design and web design.

3. Do not count on AI referrals as a replacement channel yet. With AI platforms accounting for well under 1% of web referrals in the datasets reviewed, they are worth tracking but not worth betting a marketing plan on. Showing up in AI answers is still worth preparing for, which is what AI search optimization and ongoing SEO work toward.

4. Make the website’s verification job explicit. Since customers who arrive from AI or search are often checking whether a business is real, current, and trustworthy, a site that loads fast, shows accurate details, and is free of clutter does more work than one optimized only to attract visits. A companion report in this series, AI Finds You. Your Website Verifies You., covers what that verification step involves.

The data: claims vs. observed behavior

Items marked “as reported” were obtained through secondary summaries and were not checked against the original documents. The three surveys and studies use different bases, methods, and wording and are shown together for comparison only. This report does not constitute a guarantee of results for any individual business.

A note on how Make It Loud approaches this

Make It Loud builds websites and local search strategies for North Atlanta small businesses, with an emphasis on measuring what actually brings in customers: calls, bookings, and inquiries, not just traffic. This report is part of an ongoing research series on how local customers find and choose businesses, and it is a companion to AI Finds You. Your Website Verifies You., which looks at what consumers do after an AI recommendation.

These findings inform our web design, SEO, and AI search optimization work for local businesses. This report is provided for general informational purposes only and does not constitute a guarantee of results for any individual business.