A Click Costs $1.63. The Biggest Web Game Portals Buy Almost None of Their Traffic.

Paid search charges $1.63 a click in entertainment while an ad-funded web game session earns cents, so buy distribution before you buy a catalogue. That gap is not a rounding error you optimise away with better creative or a smarter bid strategy. It is one to two orders of magnitude, it has been stable for years, and it is the reason the largest browser game destinations on the internet buy almost none of their traffic. Licensing a thousand games takes a week and a wire transfer. Finding a thousand players who come back takes longer than most licence terms run.

I keep meeting the same buyer. They have signed for a catalogue, they have a domain, they have a launch date, and when you ask where the players come from the answer is "SEO and some ads to start." The catalogue was priced to the dollar. The traffic was priced with a shrug. Six months later the portal has 400 sessions a day and a spreadsheet that only worked at 400,000.

💸 What a Visit Costs on the Open Market

Start with the number you can actually go and pay today. WordStream's 2026 Google Ads benchmarks, built from 13,474 US search campaigns running April 2025 through March 2026, put the median cost per click across all industries at $5.42. Arts & Entertainment — the category games portals bid in — is the cheapest of the 23 industries measured, at $1.63. It also carries the highest click-through rate in the study at 12.75%, which tells you the intent is genuinely there. People searching for games really do want games.

That is the good news and it is still $1.63. Search is the expensive channel by design: you are buying stated intent at auction against everyone else who wants it.

Display is cheaper per visit and worse per visitor. A 2026 compilation of display benchmarks from Digital Applied, drawing on WordStream's Q1 2026 display data and aggregated DSP reporting, puts the gaming vertical at a $2.20 Google Display Network CPM and a $4.45 programmatic CPM, with a gaming banner click-through rate of 0.95% — the strongest of any vertical they list, roughly double the 0.46% cross-industry average that Focus Digital's December 2025 study found across February–December 2025 campaign data.

Do the division. At 0.95% CTR you need about 105 impressions to buy one visit. On GDN that visit costs roughly $0.23. Programmatically it is closer to $0.47. So the honest range for a bought visit to a games portal is somewhere between 23 cents and $1.63 depending on channel, before you account for the fact that a display click and a search click are not the same person.

📊 The Two Biggest Web Game Portals Buy Almost Nothing

Now look at what the market leaders actually do, because it is public. Similarweb's July 2026 profile of poki.com reports about 193 million monthly visits, with organic search the largest single channel at 44.09%, and a keyword split of 94.57% organic against 5.43% paid. CrazyGames, at roughly 101.5 million monthly visits over the same period, is more direct-led — 48.58% of desktop visits arrive with no referrer at all — and its keyword split is 98.73% organic against 1.27% paid.

Similarweb's panel-based estimates are directional rather than audited, and the free tier only exposes the top three channels. But the direction is not subtle. The two largest destinations for browser games run audiences of 100 to 200 million monthly visits on traffic they overwhelmingly did not buy. Poki earned its way into the search results. CrazyGames turned enough of that into habit that half its visits are people typing the domain.

Poki's engagement numbers on the same profile explain why that is the only model that works: 6.51 pages per visit, nine minutes and five seconds on site, a 27.87% bounce rate. That is a returning-habit product, not a landing page. It was built over years, and years is the actual unit of cost.

🧮 Run the Arbitrage Once and You Will Not Run It Twice

Here is the arithmetic every prospective portal owner should do before signing anything. I am going to show my assumption plainly so you can change it.

The revenue side comes from realised web ad rates. AppLixir's mid-2026 network review puts web and HTML5 rewarded video at a $3.62 global eCPM and $6.98 in the United States. Other operators dispute that hard — Playgama's breakdown, relayed in Cinevva's 2026 web monetization guide, reports US web rewarded rates of $15 to $28 gross, before splits. Both can be true: one figure is a network's realised payout after splits, the other is a gross rate card. Which one you plug in changes the answer by a factor of three, which is itself a reason to run the model at both ends before committing money. If you want the splits walked through against your own traffic, that is what our monetization review exists for.

Assume — and this is my assumption, not anyone's data — that a nine-minute session on a decent portal produces three monetisable ad impressions. Then:

• At AppLixir's US rate, a session grosses about 2.1 cents. At the global rate, about 1.1 cents. At the optimistic $20 gross rate card, about 6 cents.
• A $1.63 search click therefore needs roughly 78 US sessions to break even, about 150 at global rates, or 27 even if you believe the highest published rate card.
• A $0.23 display visit needs about 11 US sessions, 21 globally, or 4 at the optimistic rate.
• Then apply your revenue share. If your licence or your ad partner leaves you 50–70% of net, multiply every one of those numbers by 1.4 to 2.

Notice what that does and does not say. Paid search for an ad-funded casual portal is not marginal — it is dead on arrival at every eCPM anyone publishes. Display in tier-1 markets is genuinely arguable if you believe the high rate cards and your retention is excellent. That is a real finding, not a slogan. It is also not a plan you should finance a catalogue purchase on, because the whole thing rests on the return-session assumption, and that assumption is the weakest number in the model.

📉 The Return-Session Assumption Is Where It Breaks

Every one of those break-even figures needs the same visitor to come back many times. Retention data says they will not.

Adjust's Gaming App Insights Report, 2026 edition puts day-1 retention across all game genres at 27% for 2025, with hypercasual and hybrid casual at the top of that range. The same report has global gaming CPI up 30% year over year to $0.56 blended, and global session counts up just 1% — costs rising into flat demand. It also notes that install trends split by region, with MENA up 2% on installs and 7% on sessions while Europe fell 7% on installs.

Day-30 is where the model dies. The most-cited genre breakdown is still AppsFlyer's Q3 2022 study of 11 billion installs across 11,000 apps, relayed in Mistplay's retention benchmarks: hypercasual retains 1.38% at day 30, casual 4.10%, puzzle 5.35%, match 7.15%. That data is four years old and I would treat the exact decimals as indicative rather than current, but nothing published since suggests the shape has improved.

And those are app numbers, from products that got an icon on a home screen and permission to send push notifications. A web portal has neither. There is no install to retain, no notification channel to re-engage with, and — as Safari's seven-day storage cap already forces on anyone paying attention — often no persistent identity to recognise the returning player by. Web re-engagement is harder than app re-engagement, not easier. If a casual app keeps 4% of its users to day 30, a browser portal buying cold clicks is not getting 78 sessions out of them.

🔁 Paid Acquisition Is a Subscription Tool, Not an Ad Tool

There is one configuration where buying traffic clears, and it is worth being precise about it, because it is where most of the durable game businesses actually live.

The problem above is that the revenue event is an ad impression worth a fraction of a cent. Change the revenue event to a recurring subscription and the arithmetic inverts. WordStream's same 2026 study puts the Arts & Entertainment cost per lead at $26.84 — the lowest of all 23 industries — against a 5.91% conversion rate. A $26.84 acquisition cost against a games subscription billing a few dollars a month pays back in months, not in impressions, and everything after payback is margin. That is the model operator game portals have run for a decade: the games are the reason to keep paying, and the billing relationship already exists.

So the rule is not "never buy traffic." It is: buy traffic only when a visit can become a recurring payment. If your portal monetises by impression, paid acquisition is a subsidy you pay to strangers. If it monetises by subscription, carrier bundle, or B2B contract, paid acquisition is a normal customer acquisition cost you can model and defend.

🔌 The Channels That Actually Clear for an Ad-Funded Catalogue

If you are not selling subscriptions, the viable traffic for a licensed catalogue has one property in common: you do not pay per visit.

• Someone else's existing audience. A publisher, a news site, a telecom deck, a bank's app, a retailer's loyalty programme. They already have the users and a reason to keep them longer. You supply the games. This is the single most reliable placement for a licensed catalogue and it is a business development activity, not a marketing spend.
• Syndication into aggregator feeds. It costs you a large share of net revenue rather than cash up front. That is a real price — but it is a price paid out of revenue that exists, not out of capital that might not come back.
• Organic search, on a multi-year horizon. Poki's 44% organic share is achievable and it is also the output of years of original games, unique pages and links. If you are shipping the same licensed titles as forty other licensees, your pages compete with forty near-identical copies for one indexed slot — budget for that, and for the time it takes.
• Embedded placements you own. White-label deployments inside an existing product — an operator or brand portal where the traffic is contractually delivered — turn acquisition from a variable cost into a fixed one you negotiate once.

The common thread: the catalogue is supply. Supply is worth nothing without demand attached to it, and the demand should be attached before the wire transfer, not after.

🧭 What to Settle Before You Sign a Catalogue Deal

1. Name the first placement. Not a channel — a named counterparty, a page, a deck slot, a contract. If you cannot name it, you are buying inventory with no shelf.
2. Write down your assumed impressions per session and your assumed eCPM. Then halve both and see whether the business still works. If it only works at the top of a disputed rate card, it does not work.
3. Model return sessions explicitly, not as a growth rate. How many times does one acquired visitor come back in 30 days? Justify the number against published retention data, and remember web has no push notification.
4. Decide whether your revenue event is an impression or a payment. This single answer determines whether paid acquisition is available to you at all.
5. Size the catalogue to the placement, not to the price list. Poki reaches nine-figure monthly visits on a curated library reported at roughly 1,500 titles. A thousand games you cannot route players to is a thousand games you are paying to host.

🌲 Where Forestry Games Fits

We have licensed HTML5 and Android games since 2017, and the buyers who do well with our catalogue are consistently the ones who arrived with a placement already in hand — an operator portal, a brand campaign, an existing app that needed a games tab, an event or retail deployment with a captive audience. The ones who struggle are the ones who bought a catalogue first and went looking for players second. That pattern is so consistent it is worth saying out loud even though it costs us deals.

If you are earlier than that, the useful next step is not a bigger catalogue. It is one conversation with whoever already has your audience: the publisher, the operator, the brand, the platform. Come back with a signed placement and a monetisation model that survives being halved, and then the catalogue question becomes easy — and small.