Zero-Click Search Hit 68%. Game Portals Are Not the Ones Who Should Panic.

Zero-click Google searches hit 68% in 2026, but game portal traffic runs on play intent and direct visits, not the informational queries AI summaries eat.

Every publisher conference this year has run the same slide: organic clicks falling off a cliff, AI Overviews swallowing the results page, an arrow pointing down. If you run a games portal, you have seen that slide and quietly wondered how long you have.

The search data is real, it is getting worse, and it mostly describes a different kind of website than yours. That is not a reason to relax. It is a reason to worry about the right things, because what will actually cost a portal its traffic over the next eighteen months is not what is on that slide.

📉 The Numbers Are Real, and Worse Than Last Year

Start with the headline. SparkToro's analysis of Similarweb clickstream data, reported by Search Engine Land in 2026, found that 68.01% of US Google searches ended without a click between January and April 2026, up from 60.45% in 2024. The share of searches producing at least one click fell 9.51 percentage points across that period — a 22.9% relative decline. Follow-up searches inside Google rose 7.2 points.

Ahrefs put a number on the mechanism. Their December 2025 dataset of roughly 300,000 keywords, compared against December 2023 through Google Search Console, found the presence of an AI Overview cut click-through rate for the top-ranking result by 58%. Eight months earlier the same team had measured 34.5%. Position two lost about half its clicks; position ten still dropped close to 20%.

The most careful study is still Pew Research Center's March 2025 browsing panel, because it watched real people rather than inferring from rank data. Across 68,879 unique searches from 900 US adults, users clicked a traditional result on 8% of visits when an AI summary appeared, against 15% when one did not. Clicks on links inside the summary itself: 1%. And 26% of users ended their session entirely after a summary page, versus 16% without.

Worth noting where these disagree. SparkToro's 68% is US-only, excludes searches inside Google's mobile app, and its authors explicitly decline to attribute the full 2024-to-2026 rise to AI Overviews. Other trackers put the zero-click rate closer to 65%. Anyone quoting a single confident figure for "how much traffic AI took" is guessing. The direction is not in dispute; the magnitude is.

🎮 Now Look at Which Queries Get Summarised

Here is the part the panic slide leaves out. AI Overviews do not fire evenly across intent, and games sit in the categories they fire on least.

Semrush tracked more than 10 million keywords through 2025. In January, informational queries accounted for 91.3% of all AI Overview triggers. By October that share had fallen to 57.1%, with commercial at roughly 18%, transactional near 14%, and navigational just above 10% — up from under 1% in January.

Read that two ways at once. The reassuring reading: "play basketball stars", "friv games", "your-portal-name" are navigational and transactional queries. They have historically been the least summarised, and there is a structural reason for it. A generated paragraph can answer "how does direct carrier billing work". It cannot play the game. When the user's goal is to do a thing rather than know a thing, the summary is a speed bump, not a substitute.

The alarming reading: navigational triggers went from 0.84% to 10.33% in ten months. That is a twelvefold increase in the exact category that protects you. The buffer is real today and it is shrinking on a visible slope. Treating it as permanent is how portals get caught.

🔖 The Big Portals Already Run on Direct Traffic

The most useful evidence is not in the SEO studies at all. It is in the traffic mix of the portals that already won.

Similarweb's estimates for Poki's regional domains put direct traffic at roughly 64% of desktop visits on poki.pl, 68% on poki.ee, and 74% on poki.ro, with organic search a distant second in each case. Those are panel-based third-party estimates for individual country domains on desktop only — not first-party analytics, and not the global picture. Treat the exact percentages as indicative. Treat the shape as instructive, because it repeats across every domain in the set.

That is what a mature games portal looks like. Search is an acquisition channel that runs once per user. Direct traffic is the business. Someone plays, comes back on Thursday, types four letters into the address bar or taps a home-screen icon, and never touches a results page again. Similarweb's month-over-month snapshots reflect a market that is moving, not collapsing: Poki up about 1.4% month over month as of July 2026, CrazyGames down about 6.6% as of June 2026. Different sites, different months, no uniform cliff.

If two-thirds of your visits arrive without a referrer, a 58% CTR haircut on the top result is a growth problem, not an existential one. If 80% of your visits arrive from Google, you do not have a portal. You have a landing page with games on it, and you were always one algorithm update from zero.

🧮 What Direct Traffic Actually Costs to Build

Nobody bookmarks a site they had a mediocre time on. Return visits are earned in the product, and the levers are unglamorous:

• Save progress without an account. Local storage plus an optional upgrade to a real account. A login wall in front of a free casual game is a return-visit tax you cannot afford.
• Make the homepage change. A grid that looks identical on day two teaches the user there is nothing new. Recently played, continue where you left off, a rotating shelf — these exist to make the second visit feel different from the first.
• Ship a real PWA install path. An icon on the home screen converts a search visitor into a direct one permanently. Most portals never prompt for it.
• Measure day-2 and day-7 return, per traffic source. Search visitors and referral visitors retain at very different rates, and the average hides it.
• Cut time-to-first-play. Every second before the game is interactive is a share of visitors who will never form a habit. This is a performance problem with a retention price tag.

None of this ranks you higher. All of it reduces how much the ranking matters.

⚠️ The Search Risk That Should Actually Worry You

While operators watch AI Overviews, the genuine threat to a licensed-catalogue portal has been sitting in Google's spam documentation the whole time.

You licence 1,000 games. You generate 1,000 pages. Each has a title, an iframe, an auto-written 80-word description, and the same five internal links. That pattern is the textbook definition of what Google's search spam policies call scaled content abuse: generating many pages primarily to manipulate rankings, with little or no value added for users. The policy has been explicit since the March 2024 spam policy update, it does not care whether a human or a model wrote the text, and Google shipped a further global spam update in June 2026 naming scaled content abuse among its targets.

This is a much sharper risk than a CTR haircut, because it is binary. AI Overviews take a percentage. A spam action takes the channel.

The defensible version costs more and works. Pick the 100 titles you will actually promote and give them pages a person wrote — what the game is, who it suits, controls, how long a session runs, what to play next and why. Let the remaining 900 live in browsable category and collection pages that earn their place through navigation rather than through 900 thin documents. A catalogue is inventory. It is not a content strategy, and search engines have got progressively better at telling the difference.

🌍 Channels That Do Not Depend on a Ranking

The strategic answer to a shrinking search channel is not better search. It is fewer dependencies.

Syndication and embeds

Placing games on other people's properties — publishers, news sites, education platforms, community sites — puts your brand in front of a session you did not have to rank for. The embed carries attribution back. This is the oldest distribution route in web gaming and it still works, because the host site wants dwell time and you want reach.

Operator and carrier placement

A games section inside a telecom app or portal is a distribution deal, not a marketing spend. The traffic is the operator's, the billing relationship already exists, and no ranking is involved anywhere in the chain.

App-store surfaces

Wrapping a portal or a curated bundle as an Android app moves discovery into store search and store editorial, which is a different algorithm with different politics. It is not free of platform risk — it swaps one gatekeeper for another — but the failure modes are uncorrelated, and uncorrelated risk is the entire point.

Owned re-engagement

Web push, email for account holders, and the home-screen icon are the only channels you control outright. They are also the ones most portals set up last, if at all.

A white-label portal deployed on a partner's domain is the same logic taken further: their audience, their traffic, your catalogue and mechanics.

📊 What to Measure Instead of Rankings

1. Share of sessions with no referrer, trended monthly. This is your independence score. If it is not rising, nothing else you are doing is compounding.
2. Day-2 and day-7 return, split by acquisition channel. Not blended.
3. Branded search volume. When people search your name specifically, you are building an asset that survives ranking changes for generic terms.
4. Games per session and sessions per user, rather than total plays. Total plays flatters a site that churns through one-time visitors.
5. Revenue per returning user versus per new user. On most ad-funded portals the gap is wide, and it is the number that justifies retention spend to a finance team.

🚫 Five Ways Operators Get This Wrong

• Rebuilding the whole content strategy around AI Overviews. If your traffic is play-intent, you are optimising for a threat that has not arrived at your door yet, while the retention work that would actually help goes undone.
• Assuming play-intent immunity is permanent. Navigational triggers went from under 1% to over 10% in ten months. Plan for the trend, not the snapshot.
• Publishing a page per licensed title on autopilot. The fastest route from a 1,000-game catalogue to a manual action.
• Buying a catalogue with no traffic plan attached. Games are inventory. Inventory without distribution is cost. This has always been true and a declining search channel makes it expensive faster.
• Reading a single vendor statistic as settled fact. The credible studies here disagree by ten points and their authors say so. Anyone selling you certainty about AI's traffic impact is selling you something.

🎮 Where a Licensed Catalogue Fits

Licensing solves the supply half of this problem — you get a deep, playable library without a studio — but it does nothing for the demand half. Forestry Games has licensed HTML5 and Android titles since 2017, and the operators who do well with a licensed catalogue are consistently the ones who treated the licence as the start of the work rather than the end of it: curation, localisation, a reason to come back on Thursday. The ones who treat 1,000 games as 1,000 ranking opportunities tend to learn the scaled-content lesson the expensive way. Browsing what players actually replay is a better starting point for a launch shortlist than sorting by catalogue size.

🧭 What to Do This Quarter

Open your analytics and find one number: the percentage of sessions in the last 90 days that arrived with no referrer. That figure tells you how exposed you are to every search change coming, and it is more informative than any industry statistic in this article.

Under 30%, you have a search dependency to fix, and the fix is retention, not more content. Above 60%, your job is to protect the habit — page speed, save states, a homepage worth reloading — and to add one channel that does not involve Google at all. In between, do both, and audit your title pages before an update makes the decision for you.

Search was never the moat. It was the on-ramp. The portals still standing in three years will be the ones that spent 2026 building something worth returning to, rather than something worth ranking.