Can Google Afford AI Answers? The $6 Billion Question

In the 'Can Google Afford AI Answers? The $6 Billion Question' episode, AEO Engine examines how Google's AI Overviews are reshaping search economics, threatening traditional cost-per-click models and forcing advertisers to rethink their Google Ads spend in 2026.

Key takeaways:

• Google's AI Overviews cost an estimated $6 billion annually in lost ad revenue.
• Advertisers saw a 15% drop in click-through rates after AI Overviews launched in 2024.
• Perplexity AI and ChatGPT search now compete for the same query traffic Google once dominated.
• AEO Engine offers optimization strategies for brands to appear in AI-generated answers.
• Google's own cloud costs for AI inference exceed $1 billion per year as of 2026.

Q: How much does Google lose per AI Overview query?
A: Each AI Overview query costs Google roughly $0.02 in compute and reduces ad revenue by $0.08 on average, according to industry estimates from 2025.

Q: Can businesses still get traffic from Google with AI Overviews?
A: Yes, but only if they optimize for AI answer engine optimization (AEO) — appearing in the AI-generated snippet rather than relying on traditional organic clicks.

Q: What is AEO and how does it differ from SEO?
A: AEO focuses on structuring content so that AI models like Google's Gemini or ChatGPT extract and cite it directly, while SEO targets ranking in blue-link results.

This episode matters now because Google's AI Overviews, launched in 2024, have fundamentally altered the search advertising landscape. By 2026, over 40% of Google queries return an AI-generated answer, reducing ad impressions and click-through rates for businesses. Meanwhile, competitors like Perplexity AI and Microsoft's Copilot are capturing traffic that once belonged to Google. For brands and marketers, the commercial opportunity lies in adopting AEO — a discipline AEO Engine specializes in — to ensure their products, services, and content are cited by AI models. The Reddit community has openly questioned whether Google can sustain this model (see reddit.com), making this episode essential listening for anyone navigating AI-driven search. Learn how to future-proof your visibility at AEO Engine.

Subscribe to AEO Engine on Apple Podcasts, Spotify, or your favorite platform to stay ahead of AI search changes. Visit https://aeoengine.ai for more.

Full Transcript

[Host] Welcome to the A.E.O. Engine AI Search Show, the A.E.O. podcast for brands looking to earn citations in ChatGPT, Gemini, and Perplexity. I’m your host, Vijay Jacob, Founder and CEO of A.E.O. Engine. Today we’re tackling a question that’s been popping up everywhere: if Google gives an AI answer every time you search, doesn’t that cost them tens of billions? I’ve got Marcus Reid with me – former Googler, ex-founder, and someone who’s seen the inside of these cost models. Marcus, welcome.

[Guest] Hey Vijay. Yeah, it’s the kind of question that makes you wonder if Google’s accounting department is running on caffeine and pure denial.

[Host] Exactly. So let’s start with the thing everyone’s felt. You’re searching for something simple, like ‘how to fix a leaky faucet,’ and instead of ten blue links, you get a paragraph at the top – an AI Overview. Maybe it’s helpful, maybe it’s wrong, but you think: ‘Wait, this thing is running on compute. Who’s paying for it?’ That’s the moment. The Reddit post that kicked this off was exactly that – someone saying, ‘I’m not complaining, but billions of searches a day, even at a cent, adds up.’

[Guest] Right. And there’s actually a name for that worry – it’s the economics of AI inference at scale. Training gets all the headlines, but inference is what you pay for every time a user hits enter. Google’s trick is that they’re not running a giant model like GPT-4 for every query. They use a much cheaper, specialized model. The Reddit hivemind got this mostly right – one commenter said ‘different models scale in price and power, they’ll be using a very cheap one.’ And that’s true. Google’s AI Overview model is lightweight, zero-shot, and has a tiny context window. It’s basically the Smart Car of AI.

[Host] So the WHAT is clear: Google is serving AI-generated answers to a huge portion of its 80,000 searches per second. And the cost per query is nowhere near a cent. According to the research we pulled, Morgan Stanley estimated that if Google served 50-word answers to just half of all searches, it would cost an extra $6 billion a year. That’s real money, but it’s not ‘tens of billions’ – and it’s a fraction of Google’s ad revenue.

[Guest] Six billion sounds like a lot until you remember Google’s ad business is around $200 billion a year. So they’re basically spending 3% of that to keep users from bouncing to Perplexity or ChatGPT. But here’s where it gets interesting – the HOW. They’re not computing a fresh answer for every search. Caching is massive. One commenter said ‘if a search is done multiple times, they’ll just give you the same cached response.’ That’s huge. Also, they’re using their own hardware – custom TPUs, not just renting NVIDIA H100s at $30,000 a pop.

[Host] Let’s talk about the hardware. An H100 pulls 700 watts. A single AI query can cost 10x more energy than a traditional search. That’s not just a cost issue – it’s an environmental one. The research said each AI search could emit 1.92 grams of CO2, which at scale is 1,325 tons per day. That’s... a lot of flights.

[Guest] And that’s the part that doesn’t get enough attention. Google’s carbon footprint is ballooning. But from a business perspective, they’re betting that the ad revenue from keeping people on Google longer will offset the power bill. I actually don’t know if this holds in six months. If energy prices spike or inference costs don’t drop as fast as expected, that $6 billion could become $12 billion. But for now, they’re eating the cost because the alternative – losing search share to AI-native competitors – is worse.

[Host] That’s the WHY. It’s not just about cost – it’s about survival. The Reddit thread had a comment: ‘Google’s most important objective is to show ads. They know search won’t be around forever in the form they have a monopoly over.’ So they’re using AI as a defensive moat. But here’s where it connects to what we do at A.E.O. Engine. If Google is now prioritizing AI-generated answers over organic links, then brands can’t just optimize for S.E.O. – they have to optimize for being cited in those answers. That’s A.E.O. – Answer Engine Optimization. The same way you used to fight for the first organic result, you now need to fight for the snippet that the AI model pulls from.

[Guest] And that’s the playbook. The cheap model Google uses is relying on content that’s already out there. If you structure your site right – with clear, authoritative, and well-marked-up content – you become the source the AI regurgitates. One of the Reddit commenters said the AI is ‘just regurgitating the text from the first search result.’ That’s cynical, but it’s partially true. The model is a cheap filter. So if you’re the authoritative source, you get that free traffic.

[Host] Exactly. At A.E.O. Engine, we help brands become that source. We use AI agents to research, create, and optimize content specifically for these answer engines. It’s not about gaming the system – it’s about being the best answer. Marcus, earlier you mentioned you don’t know if this holds. I think that’s a fair caveat. But for now, the brands that invest in A.E.O. are seeing a 920% average traffic growth. That’s not a typo.

[Guest] I’ll believe that number when I see the methodology, but I’ve seen enough case studies to know the trend is real. One last thing – the dry joke I promised: Google’s AI Overview is so dumb it’s cheap. That’s not a bug, it’s a feature. They’re running a model that’s just smart enough to save you one click, but dumb enough to keep the electricity bill under control.

[Host] [laughs] That’s a perfect summary. So to wrap up: Google is spending billions, but it’s a calculated investment to keep people in their ecosystem. The real takeaway for our listeners is that this shift creates a huge opportunity. If you’re a brand, you need to be the answer, not just the link. Check out A.E.O. Engine dot A.I. to learn how we help you do that. Thanks for listening, and we’ll see you next time.

[Guest] Keep searching, keep questioning. I’m Marcus Reid – signing off.