Tracing MFA to the source: How Amazon Ads and HUMAN are protecting the open internet

Made-for-Advertising (MFA) supply has been a known tax on the open internet for years. Thin content, heavy ad loads, inorganic traffic sourcing, sites engineered around impressions rather than audiences.

Generative AI has collapsed the time, cost and skill required to create plausible-looking publishers. Operators can now produce content at scale, fabricate author identities, clone site structures across niches and languages and construct the editorial scaffolding that makes an environment look legitimate at a glance. The same techniques used to build a credible blog can now be used to make hundreds of MFA versions in a weekend.

That’s why Amazon Ads and HUMAN are working together to detect and block evolving MFA threats and share insights that strengthen industry-wide defences.

The real cost of MFA

MFA supply doesn’t just waste ad budget. It quietly funds deceptive environments that deliver little audience value, while diverting revenue from legitimate publishers creating quality content. Every dollar that flows into these operations incentivises bad actors to create more of them, and the longer these coordinated networks go undetected, the more they erode advertiser confidence in the open internet.

Instead of asking "Does this page look like low-quality MFA?", we should ask "Who built this site and how are users ending up here?” A site can pass a casual review. Named authors, policy pages, multiple consumer categories, the visual rhythm of a legitimate publisher. But when deeper investigation surfaces synthetic identities, shared monetisation ownership and overlapping operator footprints, the calculation of risk changes.

How Amazon Ads and HUMAN raise the bar together

Amazon Ads doesn’t see supply quality protection as an add-on. Safeguarding advertiser investment is the baseline. Apps and sites are preemptively vetted for quality and integrity before becoming eligible for demand and are subject to ongoing monitoring.

To further protect buyers from evolving threats, Amazon Ads’ agentic scanning system is built to automatically identify and block suspicious supply at scale before campaigns bid on them. In 2025, this system filtered out more than 10 trillion bid requests that didn’t meet verification standards, preventing them from reaching a campaign.¹ In addition, proactive risk hunting uncovers coordinated MFA networks that no manual audit could surface alone.

HUMAN's Satori Threat Intelligence and Research team independently validates the networks Amazon identifies as MFA through indicator-based analysis. HUMAN’s methodology identifies MFA through measurable site, traffic and supply-chain indicators, instead of relying on a fixed list. Their verification spans 20 trillion interactions per week across three billion devices. Across open internet inventory, HUMAN currently flags more than 50 billion bid requests per day as MFA, underscoring the industry-wide scale of the challenge.²

These findings feed back into blocking considerations for both Amazon Ads and HUMAN. Each investigation expands the set of indicators, infrastructure patterns and operator behaviours that get filtered before reaching a campaign, so protection compounds.

When MFA is filtered out from campaigns, advertisers get more from every dollar: more efficient media spend, stronger downstream conversions and confidence that their campaigns are placed in premium environments.

What deeper investigation revealed

Here’s what one of these operations looks like up close. Amazon Ads and HUMAN recently detected and blocked a cross-niche portfolio of sites that appeared to be separate publishing properties on the surface but shared the signatures of a coordinated network:

• Dozens of sites shared the same owner as declared by ads.txt and cleared through the same monetisation path
• Sites reused legal and privacy language and pointed back to the same registered entity
• Different sites shared identical layout templates

The identity layer told the same story. Fabricated personas were found managing sites and authoring content, with profile images showing consistent signs of AI generation. The role of AI here goes deeper than content generation and manufactures the appearance of credibility.

From January to June 2026, HUMAN observed a quadruple increase in bid requests from this network, as its operators used advanced methods to conceal their activity.³ This signals how quickly the broader threat landscape is moving.

In a similar case, HUMAN used sites identified by Amazon as a starting point and mapped those signatures across the wider supply pool, surfacing additional connected sites that sat outside of Amazon’s active supply. Some were compromised legitimate sites; others were expired sites re-registered by bad actors specifically to inherit the credibility of their previous owners. These findings show how far bad actors will go to hide behaviour that could undermine advertiser trust or waste ad spend.

The bottom line

As MFA networks grow more sophisticated, the industry's standard for protection needs to evolve. Effective supply quality defence requires proactive threat detection, a combination of AI and human expertise and validation from third-party verification partners.

The operators behind these networks will keep adapting and finding new ways to look legitimate. Amazon Ads' approach is designed to evolve alongside these threats, and our customers are already benefiting from these protections. Apps and sites are preemptively vetted. Campaigns are shielded from MFA threats. And each investigation compounds into stronger protection. HUMAN’s independent validation ensures those findings go beyond Amazon’s walls and strengthens defences across the wider industry.

Sources:

¹ Amazon internal, 2026.

^2-3 HUMAN internal, 2026.