Gaming Has the Highest Invalid-Traffic Rate of Any Vertical. Budget the Clawback Before You License HTML5 Games.

Before you license HTML5 games to monetise with ads, model the invalid traffic: gaming has the highest IVT rate of any vertical, and clawbacks run late.

Ad revenue on a games portal is not final when it appears in your dashboard. It is provisional. Somewhere between thirty and ninety days after you banked it, an ad platform can decide a slice of those impressions were never seen by a person, credit the advertiser, and take the money back out of your balance. You get no line-item explanation and, on the largest platform, no appeal.

This matters more for a licensed games catalogue than for almost any other kind of inventory. Games sit in the vertical with the worst measured invalid-traffic rate. They run long sessions with high impression counts per visit, which is exactly the shape automated traffic imitates well. And the cheapest traffic available to a new portal comes from the countries with the highest IVT rates on record. Every one of those pressures pushes the same direction.

None of that is an argument against running an ad-funded portal. It is an argument for putting a number on the leakage before you sign a catalogue licence, rather than discovering it in the third month when the revenue line stops matching the traffic line.

🔢 One in Five Impressions, or One in Twelve Clicks — Both Numbers Are Real

Two credible 2026 reports measured invalid traffic and landed a long way apart. Read them together, not against each other.

Fraudlogix analysed 105.7 billion ad impressions across calendar 2025 and put the global IVT rate at 20.64% — 21.81 billion invalid impressions. Its Q1 2026 update, covering 26.3 billion impressions from January to March, came in lower at 18.12%.

Lunio's Global Invalid Traffic Report 2026, published on 21 January 2026, analysed 2.7 billion paid ad clicks across six platforms, eight industries and ten countries, and put the average IVT rate at 8.51%. Trade coverage of that report placed the resulting waste at roughly $63 billion against global digital ad spend above $740 billion.

The gap is not one of them being wrong. It is two different denominators measured from two different sides of the transaction:

• Impressions versus clicks. A bot that loads a page generates impressions cheaply and clicks rarely. Any impression-denominated sample will read higher than a click-denominated one.
• Sell side versus buy side. An impression-level supply sample sees inventory before a buyer's filters touch it. Lunio notes its data came from unprotected, monitor-only campaigns — what slips past platform filters, measured at the advertiser.
• What counts as invalid. The industry splits this into general invalid traffic — declared bots, data-centre IPs, crawlers, filterable by list — and sophisticated invalid traffic, which needs behavioural detection. Vendors draw that line in different places.

For planning purposes, the honest position is a range, not a point estimate. Somewhere between roughly 8% and 20% of what you are about to sell is likely to be judged invalid by someone, and you do not control which someone.

🎮 Gaming Is the Worst-Hit Category in the Buy-Side Data

Lunio's industry breakdown puts Gaming & iGaming at 18.49% — the highest of the eight verticals measured, more than double the 8.51% average. The rest of the table, for scale: Education and e-learning 14.41%, Telecoms and utilities 14.26%, Real estate 13.61%, Finance and insurance 10.12%, Travel 9.04%, Software and IT 6.48%, Retail 6.03%.

Two caveats before you paste that 18.49% into a model.

First, the category bundles gaming with iGaming, and Lunio attributes the rate largely to gambling's economics — high-value conversions, aggressive CPC competition and sign-up bonuses. A casual HTML5 arcade is not a sportsbook. Your true rate is probably lower than 18.49%, and nobody has published the number for casual web games specifically.

Second, this is click data from advertisers buying traffic. As a portal operator you sit on the other side of that trade for your ad revenue — but you sit on this side the moment you buy traffic to fill the portal. If you run paid acquisition for a games site, that 18.49% is a number about your media budget, not about your inventory.

The uncomfortable version: a portal that buys gaming traffic and sells gaming impressions is exposed to invalid traffic twice, on both sides of the same visit.

🌍 The Cheapest Traffic Sits Where the IVT Rates Are Worst

Fraudlogix's regional split for 2025 tracks almost inversely with CPM. Asia-Pacific ran 27.85%, the United States 23.69%, Latin America 17.90%, MENA 13.78% and Europe 7.80%.

The Q1 2026 country detail is sharper still. Indonesia measured 59.17%, Ukraine 53.35%, Australia 32.55% and South Korea 23.47%. At the clean end: the Netherlands 1.02%, France 1.19%, Germany 1.99%, the United Kingdom 3.25%. The United States, 38.5% of that quarter's sample, came in at 20.37%.

Sit with the Indonesia figure for a moment, because it collides with a decision plenty of operators are making right now. Indonesia is the volume story in Southeast Asian casual gaming, and traffic there is cheap. If well over half the measured impressions in that market carry invalid-traffic signals, a portal priced on raw session volume in Indonesia is not priced on anything that will pay out.

Two more splits worth carrying into a plan. Device: desktop ran 27.03% in 2025 against mobile's 19.30%, and by Q1 2026 the two had converged — desktop 18.60%, mobile 18.16% — which Fraudlogix flags as a first in its data. Browser: Chrome 21.75% against Safari 13.24% across 2025. A portal whose audience skews desktop-Chrome in Asia-Pacific and one whose audience skews mobile-Safari in Western Europe are not running the same business, whatever the traffic dashboard says.

⏳ You Bank the Money in March and Lose It in June

The mechanics of the clawback are where this stops being an industry statistic and becomes a cash-flow problem. Google's own Ad Manager documentation on deductions from earnings is unusually blunt about it:

• Google credits advertisers for invalid traffic accrued 60 days prior to the date of invoice.
• Once it starts processing those credits, it may take 30 days or more for the deduction to show up in a publisher's account.
• You cannot appeal the deduction. Google states it will not disclose the specifics of what it detected, to protect the integrity of its detection system.

Add those windows together and a deduction can land roughly ninety days after the traffic it relates to. For a portal, that is a quarter — long enough that you have already reported the revenue, already reinvested it in acquisition, and possibly already used it to justify expanding the catalogue.

The operational consequences are specific:

• Do not treat month-one eCPM as a run rate. Give any new traffic source a full quarter before you extrapolate from it.
• Hold a reserve against provisional revenue. Operators who have been through this typically carry a percentage of ad income as unrecognised until the deduction window closes. Pick a figure from your own history, not from a report.
• Do not fund a fixed licence commitment out of unreserved ad revenue. A catalogue fee is due on a date. Ad revenue that can be reversed ninety days later is not the right instrument to cover it.
• Understand the tail risk. Repeated or severe invalid-traffic problems can end in account termination, at which point the question is not what you earn but whether you have a second demand source at all.

🖼️ Whose Domain Is on the Ad Request?

Here is where the licensing decision starts to matter mechanically rather than commercially. Ad requests report the page and domain making them. The IAB Tech Lab's authorised-seller standards — ads.txt, app-ads.txt, sellers.json and the supply-chain object — all key off that reported domain to establish who is allowed to sell the inventory.

A licensed HTML5 game usually runs in a cross-origin iframe. If that iframe is served from the licensor's domain and the ad calls originate inside it, the ad stack is operating from an origin whose ads.txt file you do not control. Whether the top-level page or the frame's own domain ends up as the domain of record depends on how the tags are configured, and the standard advice in ad ops is the same either way: serve ads from a domain whose ads.txt you own.

That turns three delivery models into three genuinely different monetisation positions:

• Licensor-hosted embed. Fastest to launch, least code. The ad stack, the origin and the seller chain are somebody else's, and typically so is a share of the revenue. Fine when the games are a feature of a larger site rather than the business.
• Self-hosted builds on your own origin. You take delivery of the files and serve them from your domain or a subdomain of it. Your ads.txt governs, your sellers.json entry is the one buyers verify, and your traffic-quality record is yours to defend. This is the position an ad-funded portal should be aiming for.
• Source-level delivery. Adds the ability to change the ad integration inside the build — swap an SDK, remove a network, add a consent path. Worth it when you are running a large catalogue on a single mediation setup.

Settle this in the licence rather than after the first deduction. The clauses to name explicitly: hosting rights and permitted origins, who owns the ad slots inside the build, whether you may replace the ad SDK, and whether the licensor's own analytics or ad calls fire from your pages. Our notes on portal monetization go into the revenue-side arithmetic in more detail.

🚫 Six Ways a Games Portal Manufactures Its Own Invalid Traffic

Not all of the leakage comes from outside. A meaningful share of what gets deducted is self-inflicted, and all six of these are fixable without spending anything.

• Aggressive refresh on long sessions. Games produce ten- and twenty-minute sessions. A short refresh interval turns that into an impression count that looks synthetic even when the player is real.
• Prefetching game pages. Speculative preload of the next title's page can fire ad calls for a page nobody ever looks at. Those are unviewed impressions by definition.
• Incentivised or exchange traffic. Traffic swaps, autosurf networks and "play for points" acquisition deliver sessions that ad platforms are specifically built to identify.
• Uptime monitors and internal QA hitting live ad slots. Synthetic checks that load the production page every minute from a data-centre IP are, from the platform's side, indistinguishable from a bot. Exclude your own monitoring and staff IPs.
• Cheap paid traffic bought on volume. Pop and redirect inventory is priced the way it is for a reason. Buying it to hit a session target imports someone else's fraud problem into your account.
• Leaving the catalogue crawlable in ways that load ads. Every licensee of a shared catalogue has near-identical game pages. Crawlers hit them hard, and a crawler that executes your ad tags is a general-invalid-traffic impression with your name on it.

📋 What to Pull Before Your Next Traffic Buy

A short, unglamorous audit that most portals have never actually run:

1. Your last six months of ad revenue with deductions shown separately, expressed as a percentage of gross. That is your real leakage rate — the only one that describes your site.
2. Deduction rate split by traffic source. One channel is usually responsible for most of it.
3. Deduction rate split by country, held against the CPM you earn there. Cheap traffic that is 30% deducted is not cheap.
4. Impressions per session by title. Outliers point at refresh settings or a broken build, not at engagement.
5. Which domain your ad calls report, checked in the network tab, and whether that domain's ads.txt lists your seller accounts.
6. Whether your monitoring, QA and office IPs are excluded from ad serving. Confirm it rather than assume it.

🎯 What to Settle in the Licence When You License HTML5 Games for an Ad-Funded Portal

Licensing direct from the company that holds the catalogue means the delivery questions above are answerable in the same conversation as price. Forestry Games has operated since 2017 and licenses a catalogue of 1,049 titles spanning HTML5 builds and Android APK builds, with titles also published on Google Play and the Apple App Store, and develops HTML5 games in house.

For an operator whose revenue is ad-funded, the licence terms that matter are the ones that decide where the game runs and who controls the ad slots: HTML5 builds you can host on your own origin, APK builds where the distribution route calls for them, source-level delivery where you need to change an integration, branding and white-label options, and hosting arranged either way. A catalogue of that size also lets you match the mix to the market you can actually monetise cleanly, rather than taking everything and sorting it out later. You can browse the catalogue or license HTML5 games by scope, and ask for a licence scope written around the hosting model you need.

🧸 Licensing Branded Games When the Money Is Not Ad-Funded

The clean way out of the invalid-traffic problem is revenue that does not depend on impressions at all — a campaign paid for by a brand, a carrier bundle, an event activation, a subscription. Branded content is the usual route to it.

Forestry Games works with branded IP and has brand partnerships including Disney, Nickelodeon, Cartoon Network and Warner Bros. Businesses can license branded game content through Forestry Games for campaigns, portals, events and apps, alongside the generic catalogue. That is a capability statement and nothing more: terms, availability and approvals are set case by case, and a branded licence carries its own clearance and approval process.

If a branded activation or a subscription tier is the direction, the next step is a scope conversation rather than a download — tell us the surface, the territory and the window, and ask for a portal demo or a membership and licensing option that fits it.

✅ Put a Number on It This Week

The single most useful thing in this post is not a benchmark from a report. It is the deduction line in your own ad account, expressed as a percentage of gross revenue, split by traffic source and country. Published rates range from 8.51% to 20.64% depending on who is counting and what they count; your number is the only one that will appear on your P&L, and most operators have never looked at it directly.

Pull it before you commit to the next traffic buy or the next catalogue term. If the leakage is concentrated in one source, stop buying that source. If it is concentrated in one country, reprice that country. And when you next license HTML5 games, put the hosting origin and the ad-slot ownership in the term sheet on the first pass — those two clauses decide whether the traffic-quality record you are building belongs to you or to somebody else.