The fight against the malicious ads on the internet The advertising landscape has taken a significant turn with the introduction of Gemini, Google's artificial intelligence model that has now been fully integrated into its advertising safety systems. The company claims that, thanks to this technology, it can block virtually all harmful ads before they reach users.
In its latest ad security report, Google details that by 2025 its systems, powered by Gemini, They intercepted more than 99% of the ads that violated the policies before they were even published. We're talking about billions of attempted fraud and scams that never appeared on the screens of those who browse the internet daily, including in Spain and the rest of Europe.
A growing problem: the rise of malvertising and advertising scams
The call malvertising (malicious advertising) For years, cybercriminals have been making life difficult for users: ads that seem normal but redirect to fake websites, install malware, or try to steal banking information. With the rise of generative AI, cybercriminals have professionalized these techniques and are able to launch large-scale campaigns in a matter of minutes, even exploiting new advertising spaces like those that WhatsApp ads are being tested.
According to data made public by Google, by 2025 They blocked or removed more than 8.300 billion ads that violated the platform's policies. In addition, around 24,9 million advertiser accounts for seriously violating the rules of use.
Within that volume, economic fraud accounts for a particularly significant portion: more than 602 million ads linked to scams And more than 4 million accounts directly linked to fraudulent activity were suspended. In other words, these are not isolated cases, but rather a large-scale digital fraud industry.
This trend is noticeable worldwide, including in the European and Spanish markets, where it is becoming increasingly common to find seemingly legitimate campaigns which are actually trying to impersonate banks, online stores, or investment services. To stem this tide, old filters based solely on keywords have clearly fallen short.
From filtering words to understanding intentions: what Gemini brings to the table
For years, the detection of problematic ads relied on lists of suspicious terms and fixed patternsThe attackers quickly learned how to bypass these filters: small changes to the text, slightly different images, or formatting tricks were enough to infiltrate the systems.
Gemini changes its approach. Google's models analyze hundreds of billions of signals to understand the intention behind each ad and each advertising campaign. They don't just focus on the visible content, but cross-reference data such as account age, historical behavior, campaign structure, and posting patterns; at the same time, this evolution competes with multimodal AI models deployed by other companies.
This approach allows us to identify whether an advertisement is more like a legitimate offer or a sophisticated hook, even when designed to appear completely normal. According to the company, this finer level of analysis has allowed them to reduce by a 80% of suspensions were incorrect from advertiser accounts that did comply with the rules.
In practice, this means that companies that advertise legally—including Spanish and European businesses that rely on online advertising—are less likely to see their campaigns blocked by mistakewhile strengthening the filter against scammers trying to sneak in among legitimate ads.
This ability to understand context and intention also makes it easier to distinguish between sensitive categories, such as financial services or certain dating platforms, where Gemini must separate genuine offers from deceptive or potentially abusive proposalsThis is especially sensitive in jurisdictions with strict regulations such as the European Union.
Real-time blocking and less exposure to risk
One of the key points of the new system is the response speedA single malicious ad can reach thousands of users in a very short time, so every minute counts. This is where Gemini has become a central piece of the security apparatus.
Google explains that, by the end of 2025, most of the Responsive Search Ads Ads created through Google Ads were already reviewed almost instantly. Harmful content was blocked the moment the advertiser submitted it, before it even appeared in any search results.
The company plans to extend this real-time review to more advertising formats, including visual ones, over the coming months. If this rollout is completed, users' exposure—including in Spain—to potentially dangerous ads could be reduced even further, leaving less room for so-called "zero-day" attacks, which exploit vulnerabilities or newly emerging trends.
This massive automation frees human teams from much of the mechanical work and allows them to focus on the more complex cases, where it is necessary expert opinion and detailed analysisAccording to the report's data, in 2025 Google's security teams acted on more than four times the number of user reports compared to the previous year.
In parallel, artificial intelligence-based systems helped to intervene in more than 467 million pages out of a total of more than 480 million studied, with detection rates exceeding 97% in that area. In other words, the AI ​​doesn't just look at the isolated ad, but also the environment it targets.
AI vs. AI: How Defenses Respond to New Frauds
The current scenario is a kind of digital arms race: the Cybercriminals rely on generative AI to create highly polished ads, compelling copy, and fake websites that are difficult to distinguish from the real ones, while platforms are forced to deploy equally advanced models to keep up.
Gemini is positioned right in the center of that battle, using neural networks capable of identifying coordinated campaign patternsAdvanced phishing or identity theft techniques, as well as subtle changes designed to evade automated controls, are used. The goal is to raise the bar to the point where maintaining large-scale fraudulent operations becomes much more costly and complex.
Data from the last year suggests that this approach is beginning to bear fruit: with more than 99% of malicious ads stopped before being publishedThe scope for malvertising campaigns is considerably reduced. Even so, Google acknowledges that some cases still manage to slip through, which is why it maintains other monitoring mechanisms.
For Spanish and European users, this translates, in theory, into a a cleaner environment when searching for information, shopping online or hiring servicesReducing the number of visible fraudulent ads lowers the chances of falling victim to financial scams, malicious software downloads, or websites that impersonate banks, stores, or government agencies.
Nevertheless, European authorities and regulators remain very attentive to how these automated systems are implemented, both because of their impact on security and their influence on the advertising ecosystem and the visibility of small and medium-sized enterprises that depend on these platforms; the discussion also encompasses issues regarding Privacy and advertising on iOS and its effect on the advertising market.
More layers of defense: advertiser verification and strict rules
Although artificial intelligence carries the weight of daily detection, it is not the only barrier that has been strengthened. Google maintains a advertiser verification program which forces many accounts to verify their identity and country before they can publish ads.
This verification works like a first line of defenseIf a malicious actor fails the process, they don't even gain access to the system. According to the company, this additional filter helps build trust in the ads that are ultimately displayed by providing information about who is truly behind each campaign.
Furthermore, very strict policies are applied to risk categories, ranging from malware and deceptive content to techniques expressly designed to circumvent controls. In the most serious cases, penalties include immediate suspension of accounts and permanent ban to be announced again.
The report also mentions that, throughout 2025, They blocked or restricted more than 480 million web pages Measures were taken against more than 245.000 publisher sites. This cleanup affects the entire advertising network and reduces the available space for placing malicious ads.
All of this is combined with improved management of user complaints. With Gemini's help, security teams can process reports more quickly and escalate urgent cases, allowing them to remove potentially dangerous campaigns faster once they pass initial screening.
Overall, the integration of Gemini into Google's security systems represents a significant leap forward in the ways to stop malicious adsThe approach shifts from reacting after the fact to proactive blocking, relying on massive signal analysis and interpreting the true intent of each campaign. For those browsing from Spain and Europe, this should translate into a slightly less hostile environment, with fewer visible traps and a more discreet, yet crucial, role for artificial intelligence behind the scenes.
