South African advertisers may be losing between R2.8bn and R12bn a year to programmatic ad fraud. The tools to stop it exist. What is missing is anyone with a reason to care.
Most older houses have a damp patch somewhere. A faint stain on the bedroom wall, a musty smell every winter. The owners know what it means: water is seeping in behind the plaster, slowly and out of sight. And most of them do the same thing: they paint over the stain, because opening the wall is expensive and messy, and might reveal something worse.
That is how many South African advertisers treat digital ad fraud in 2026. Ask a marketing team whether some of their spend reaches bots, and they will readily agree. Ask for their own figure, and the conversation usually stalls. They acknowledge the problem in general but ignore it in practice because finding it is inconvenient and fixing it seems impossible. Meanwhile, the damp keeps spreading.
How bad is the leak?
I recently gave our working estimate to ITWeb’s Nicola Mawson. South Africa's digital advertising market is worth roughly R35bn to R40bn a year, and programmatic buying accounts for an estimated 40% to 50%. Across the local campaigns we have analysed, programmatic wastage has commonly ranged from 20% to 60%. Applied together, the loss falls somewhere between R2.8bn and R12bn a year.
That range may seem broad, but even the low end is concerning. I can’t claim our sample represents the entire SA industry, and reliable, industry-wide fraud figures don’t exist. What I can say is that those results are typical of what we regularly encounter in our work with large companies. Independent data confirms we're slightly worse off by world standards: Fraudlogix's 2026 analysis of 105.7 billion impressions found South Africa's invalid traffic rate at 22.94%, above the global average of 20.64%.
The more useful number comes from one of many live campaigns we are forensically auditing for a global advertiser. Of millions analysed, 41% are currently confirmed fraudulent and 12% confirmed legitimate. We only call an impression fraudulent if it triggers at least three separate warning signs, so we leave the remainder unresolved rather than assuming guilt. Then the encouraging part: as the campaign was optimised, confirmed fraud in the most recent 20,000 impressions fell to 27%. We can identify the leak and slow it.
This is not waste. It is theft
The industry prefers the word ‘wastage’ because it sounds accidental. Dropped connections, a clumsy media plan, bad creative. But that’s not it; we are describing the deliberate, systematic diversion of advertisers’ money by criminal means.
The techniques are mundane:
- Several ads stacked behind the one you can see
- Ads served into spaces zero pixels wide
- Ad calls generated by code never meant to show anything to anyone
- Fake sites spun up overnight, claiming to be legitimate news sites and harvesting clicks
Each is billed as an impression, yet none had a realistic chance of reaching a human being. The raw material is cheap, too: ITWeb counted more than 2,200 website-traffic services on Fiverr, with some promising thousands of visitors for as little as $5.
This is not new. US Senator Mark Warner told the Federal Trade Commission that he had warned about bots in digital advertising a decade ago. Ten years is a long time to keep painting over the same stain.
An open secret?
Corrective action, or ‘mitigation’ is possible. Integral Ad Science's own 2025 research found fraud rates of 10.9% in campaigns without mitigation, compared with 0.7% in those with mitigation. Hagai Shechter, chief executive of Fraudlogix, put it neatly: “Ad fraud isn't a technology problem any more.” He means it has become a human problem, one of incentives and convenience.
In most agencies, ad tech managers are under pressure to spend their client budgets, which means hitting volume goals. If fake ad impressions are removed, volumes are reduced, which means that revenue is lost.
The awkward truth? Most agencies use legacy tools built for a simpler era. Programmatic has evolved to a level of complexity these tools simply can’t track.
It’s a structural problem: the platforms sell the inventory and then provide the measurement that vouches for it. When a dashboard shows low fraud figures, everyone who signed the contract takes comfort. But that low number measures what the filter looks for, not what is actually there. If you go into the garden looking for lizards, you will catch lizards, not every reptile out there.
Matt Sanders, formerly a senior PHD executive, argued in The Media Leader that fear has drained courage from agency life. Ad fraud survives on the same fear. Nobody forbids the question. Everybody simply understands the agenda.
In 2024, Adalytics reported that one widely used filtration product classified declared bots as valid human traffic 17% of the time and undeclared bots 77% of the time. The vendors disputed the research, and that argument is still ongoing.
But the fact remains, and our experience bears it out: ad fraud is far greater than the industry is prepared to accept. Many respected publishers don't realise fraudsters are using their sites to ‘hijack’ ad revenue.
The silence now has an expiry date
Until now, not knowing has been the cheapest position to take. It costs nothing, and no one has ever been asked to account for it. That is about to change.
ISA 240 (Revised), the auditing standard on fraud in financial statements, applies to periods beginning on or after 15 December 2026. It sharpens the expectation that auditors will challenge management's assumptions and weigh contradictory evidence. Digital media is often one of the largest discretionary lines in a listed company's accounts, invoiced by the parties being paid and confirmed by the parties delivering it. Sooner or later, someone in an audit committee meeting will ask how the company knows those impressions reached people.
The reward for getting it right is real. The ANA's Q4 2025 programmatic benchmark found that well-governed advertisers (those who audited their supply chain) turned 56.7% of programmatic spend into quality impressions, compared with 37.5% for the rest (those who relied on legacy reporting systems).
That’s quite a gap!
Someone has to open the wall
Fixing this doesn't require a new tech stack or a public confession. It needs a second measurement, at impression level, running alongside the first on your own campaigns. That is the business my company is in, so read this as an informed opinion.
We ask nobody to take our word for it: a simple test audit will suffice. If our numbers agree with your dashboard, you have spent a fortnight and gained a defensible answer. If they don't, you have found your share of the R2.8bn to R12bn, on your own timetable, before someone with an audit mandate finds it for you.
Another coat of paint will not dry a damp wall, and a more comprehensive solution is not as impossible as it seems: in the campaign previously mentioned, confirmed fraud fell from 41% to 27% after we traced the source of the leak.
If you would like a free forensic audit of your campaigns, get in touch, and we will put the numbers side by side.
