Industry executives and experts share their predictions for 2026. Read them in this 18th annual VMblog.com series exclusive.
By David S�n�cal, Director of Engineering, Fraud and Abuse at Akamai
A year ago, I predicted that bots would continue evolving and the most advanced would continue challenging the effectiveness of bot management products. We made significant progress advancing our bot-detection engine, but, as I predicted, the problem is far from solved. What I had not expected, however, was the explosion of agentic traffic and the emergence of a multitude of new protocols designed to enable Agentic commerce, content monetization, and authentication of AI agents. Users are increasingly interacting with the web through their trusted AI agents, rather than through a traditional web browser or mobile app. This shift in user behavior challenges the established bot management models and strategies, where it is no longer only about detecting humans vs. bots but also about adding AI agents to this already complex equation, assessing their intent, and detecting abusive behavior.
In 2025, we saw the emergence of several protocols, such as web-bot-auth, to authenticate bots and agents. Know Your Agent (KYA), supported by the established Know Your Business (KYB) and Know Your Customer (KYC) models, also emerged to authenticate the individuals and/or businesses behind agents and “good bots”. KYAPay, Visa’s Trusted Agent Protocol, Mastercard’s Agent Pay, and the Agent Payment Protocol (A2P) are other protocols designed to enable agentic commerce. Whereas to support the media and publishing industry, the Real Simple Licensing (RSL) standard aims to help establish licensing agreements between content owners and AI agents. This year saw the introduction of new protocols and industry standards. But ultimately, these protocols will only matter once they get adopted by AI platforms. In 2026, we can expect or hope for some consolidation between the proposed standards to make the implementation and adoption easier for the industry. Bot management products will evolve to help simplify and obfuscate the complexity of these protocols, enabling easier adoption by merchants and website owners. The new authentication protocols, when adopted, will provide a more accurate means to authenticate and categorize bot traffic and provide more actionable insights for website owners using these products.
From the evasive bot perspective, their sophistication will continue to advance, exploiting detection blind spots or safeguards meant to prevent false positives. In 2026, bot detection engines need to become more aggressive. More aggressive detection methods sometimes lead to higher false-positive rates. To counter this, bot management products will need to combine their most aggressive detection methods with a secondary detection layer that can challenge users exhibiting anomalous fingerprints or behaviors, enabling deeper assessment and more accurate decision-making. The detection engine will need to be intelligent enough to decide the type of challenge, transparent (no user interaction), or a challenge that encourages user interaction (CAPTCHA), to allow the collection of the correct type of data to help the detection engine in making an accurate decision. Unlike the traditional model, where security professionals would choose a response strategy (deny, challenge, tarpit) for all evasive bots, more advanced products will become more intelligent, deciding what next step to take to either refine their decision or outright block the traffic, ultimately yielding better accuracy and less friction with end users.
In 2025, bot detection became more nuanced by introducing the notion of intent. Bots are used for a range of purposes, including credential stuffing, account-opening abuse, scalping, and simple data collection through scraping. The intent behind credential stuffing is clearly fraud and account takeover, while account-opening abuse and scalping tend to go together, and the legality of such activity is clearly challenged in some countries, like the US, under the BOT Act. Finally, scraping is a lot more nuanced in terms of intent. For the most part, the intent is commercial use for competitive analysis, while in some cases it could simply support public safety, societal research, opinion analysis, or support investment strategies. This is the age-old story: businesses need market and competitive data to define their strategy, but no one is ready to share theirs. I expect the concept of intent to evolve in 2026, providing more context for web security professionals to decide how to respond to specific bot activity.
##
ABOUT THE AUTHOR
David S�n�cal is the Director of Engineering, Fraud and Abuse, at Akamai and is the author of The Reign of Botnets (2024). He is passionate about improving internet safety and is an expert in online fraud and bot detection.
