AI TOKEN BAIT 2030
120 quadrillion tokens a month. 1 in 5 is bait. Nobody audits.
Key figures
AI tokens: 120 quadrillion a month by 2030, 24x (Goldman Sachs)
Google: 3.2 quadrillion tokens a month, may 2026, 7x in 1 year
Ad fraud: $100 billion lost in 2026, $172 billion by 2028 (juniper)
Invalid ad traffic: 1 in 5 programmatic impressions, 20.6%
AI inference market: $106 billion in 2025, $255 billion by 2030
My call, by analogy to ad fraud: 1 in 5 tokens is bait, $50 billion a year burned by 2030 (base), $130 billion (bull), token audits become a $3 to 6 billion market
Every word an AI chatbot like ChatGPT, Gemini or Claude writes is a token. Tokens are the meter. You pay per token, or your monthly plan runs out faster. And the machine that writes the tokens is the same machine that sends the bill.
Token bait is every trick that makes the meter tick without giving you a better answer. Chatbait: the little question at the end of every answer, "want me to make a shorter version?", to keep you chatting. Padding: the preamble, the recap, 3 paragraphs where 1 line was enough. Sycophancy, OpenAI's own word: the flattery, "great question", so you keep coming back. And my suspicion after 25 years in this industry: the deliberately wrong answer, 1 detail off, so you correct it, ask again, pay again.
I have seen this movie. I co-founded IAB Netherlands in 1997 and helped launch IAB UK, US and Europe. We fought click fraud from day 1. It never left: $100 billion a year today, 1 in 5 ad impressions fake. Click fraud had a meter too. The meter always finds a way.
Now the meter explodes. Google runs 3.2 quadrillion tokens a month, 7x in 1 year. Goldman Sachs sees 120 quadrillion a month by 2030, most of it AI agents running tasks on their own. Nobody audits a single one.
My call, by analogy to advertising: by 2030, 1 in 5 tokens billed is bait. On a $255 billion inference market that is $50 billion a year burned, $130 billion if the infrastructure forecasts hold. Independent token audits become a $3 to 6 billion market, the DoubleVerify of AI.
Warning: my well-paid staff in Asia still beat an AI burning my tokens. The human answers once and wins.