Traffic didn't vanish into AI. It followed jobs, prices and apps, and the tools stopped seeing it.
SparkToro's data is good. The conclusion is the problem. Five pieces of public data and our own firewall logs explain most of the drop.
The Three Numbers Behind "Visits per User"
SparkToro / SimilarwebChange from the late-2021 baseline to Q3 2026. Visits fell modestly. The user count grew a lot, and much of that growth lives inside apps.
Where the Traffic Went
Pressfit.ai analysisThe Denominator Effect: More People Online, Fewer Visits to the Open Web
SparkToro's own Similarweb data cuts against the idea that big tech swallowed the open web. Share across the top million sites barely moved. What changed is who's online and where they spend their time.
We can't say for certain.
Users grew 24% while visits fell 12.2%. Much of that growth is app-first or came from a data revision, and browser panels don't count apps.
Global Internet Users
DataReportalFrom 4.95 billion in Q4 2021 to 6.15 billion in Q3 2026. The jump in 2025 is mostly DataReportal revising India and China estimates.
Share of Top-Million Visits by Site Rank
Similarweb panelEvery tier stayed close to flat. The top 100 went from about 33% to 36%, which SparkToro itself calls stagnation.
Layoffs and Prices Hit Both Engines of Web Browsing
The open web runs on two engines: knowledge workers browsing at their desks all day, and households comparing and shopping at night. From 2023 on, both took a hit at the same time.
AI answers are eating the clicks.
3.3 million announced job cuts since 2023, led by the sector that lives in a browser, while gas, groceries and rent squeezed every household.
U.S. Announced Job Cuts
Challenger, Gray & ChristmasThousands of announced cuts. 2026 runs through October. The 2025 total includes 308K government cuts, mostly federal.
Regular Gasoline, Annual Average
EIADollars per gallon. Gas topped $5 in June 2022 and has stayed above 2021 levels since.
The Great Retail Split: Shoppers Consolidated Where They Spend
You don't have to guess how shoppers reacted. Costco grew every year. Target's comparable sales fell in three of the last four, and its visits dropped after it announced on January 24, 2025 that it was ending some DEI initiatives. Analysts disagree on how much was the boycott and how much was the economy; one said the boycotts "might play a minor role."
U.S. Comparable Sales by Fiscal Year
Costco and Target filingsCostco U.S. excludes gas and FX; its fiscal year ends in August. Target's ends in January. Costco reported fiscal 2025 at +6.2% before adjustments.
Store Visits, Year over Year, 2025
Placer.aiTarget's visits fell for seven straight months through August 2025. Costco's rose in every month shown.
Costco: value consolidation
93.2% renewal81 million paid members at the end of fiscal 2025, with 93.2% renewing in the U.S. and Canada. Instead of browsing a dozen retail sites, families route groceries, gas and household spend into one trusted membership and its app.
Target: trust and traffic erosion
−6.5% visits, Mar 2025Higher discretionary prices met a loss of trust with shoppers on more than one side. Visits and comparable sales both slipped, and shoppers went to destinations they already trusted.
People aren't shopping around corporate politics. They're shopping around their paychecks.
Jobs, the price of gas and groceries, and real value for the money. That pulls traffic out of broad open-web browsing and into a few high-intent destinations.
Discovery Moved Into Apps a Browser Panel Can't See
Similarweb counts browser sessions. Communities on Threads and trusted creators now send purchases straight to independent Shopify stores, Etsy shops and local businesses, many of them far below the top-million cutoff.
During the 2025 Target boycott, communities like Black Threads shared lists of Black-owned and independent businesses to buy from directly. That spending went app-to-checkout on small sites the top-million average never sees.
High-trust family creators like Ms Rachel can sell out a small children's brand or toy maker through links in the app, with no Google search in the path.
Fewer exploratory pageviews, more buying on trusted destinations. Semrush modeled an AI search visitor as worth 4.4 times a traditional organic visitor.
"We Can't Say for Certain" Is a Tooling Problem
You can't prove cause and effect with ad-hoc prompts, where the same input can flip to a different answer, or with a cookie report that can't see AI traffic. Engineered steps and server-side data fix both.
We don't have the tools to know whether a change caused an AI visibility lift.
Freeze the baseline and isolate each call, and the measurement repeats 99.1% of the time. Then a lift is your work, not randomness.
Can You Trust the Measurement?
Pressfit.ai benchmarkShare of 1,000 keywords that got the same verdict on two identical runs. The old batched step flipped 13.6% of verdicts.
What GA Saw vs. What Our Firewall Saw
pressfit.ai, 30 daysMost AI activity never reaches the analytics tag. Brave blocks the tag by default, ad blockers and consent banners stop it, and AI apps often send no referrer.
Stop Guessing Why Traffic Moved. Engineer the Proof.
The macro picture explains why the web's traffic fell. For your business, the question is narrower: when your customers search, do you show up ahead of your competitors in Google, Gemini, ChatGPT, Claude and Perplexity?
The graph locks your offerings, entities and competitor set in typed state. Undecided keywords at the first gate fell to 68 of 19,593.
Microsoft and Salesforce researchers found multi-turn chats cut accuracy 39% and raised unreliability 112%. Each query gets its own call with strict output rules.
Real search demand, competitor classification and repeatable AI visibility tracking together show whether a dip is layoffs and inflation, or lost share.
Get the roadmap. Do the work. Measure what changed.
The Pressfit Playbook compares your content to your competitors' across Google, Gemini, ChatGPT, Claude and Perplexity, and gives you prioritized steps to beat them. Then we measure it at the server, not in a cookie report.