When Intelligence Is Metered Like Electricity, SEO Changes: Three Scenarios for the Next Five Years

In March 2026, at BlackRock’s Infrastructure Summit, Sam Altman described OpenAI’s endgame in one sentence: “We see a future where intelligence is a utility like electricity or water, and people buy it from us on a meter and use it for whatever they want to use it for.”

It is easy to read that as just another abundance slogan. But sit with the mechanism for a second, because it has a consequence almost no one in SEO is talking about: if intelligence is metered, then answers have a marginal cost. Today, a search result is “free” attention, Google shows ten blue links and the click costs nothing to produce. In a metered world, every AI answer that names your brand is the engine spending compute to choose you. That single shift reorders the incentives underneath search.

This is a prediction piece, so treat it as one: a set of scenarios and bets, not forecasts I can prove. But each scenario is anchored to something already happening, and the practical advice at the end holds regardless of which one wins.

The premise: metered intelligence

Altman’s framing is that AI providers are, functionally, selling tokens. You pay for the reasoning you consume, the way you pay for kilowatt-hours. He even borrowed the old nuclear-era line, “too cheap to meter”, to signal the ambition: make intelligence so abundant the price approaches the cost of the compute behind it.

Whether it gets “too cheap to meter” or stays expensive is the fork in the road. Both branches change SEO, just in opposite directions. Here are the three scenarios I think are live.

Scenario A: Abundance, answers everywhere, clicks nowhere

If intelligence gets genuinely cheap, AI answers stop being rationed. Every query, on every surface, gets a synthesized response. We are already most of the way here: in the first four months of 2026, roughly 68% of Google searches ended without a click, and on searches that trigger an AI Overview that figure climbs to about 83%. In AI Mode it is 93%. Position-one click-through for informational queries with an AI Overview has collapsed from 7.6% to around 1.6%.

In the abundance scenario, that trend doesn’t reverse, it becomes the entire weather system. The click is no longer the unit of value. Being the cited source is.

What SEO becomes here:

• Measurement moves from clicks to citations and mentions. “Did we rank?” becomes “Were we named, and were we named as the authority?”
• Content depth becomes a compute-worthiness signal. Cheap intelligence will still ration attention, the model picks the few sources worth synthesizing from. Thin, derivative pages don’t earn that selection.
• Brand search becomes the scoreboard. When the answer is the destination, the downstream signal you can still measure is whether people later search for you by name.

Scenario B: Premium tiers, a cognitive divide

The opposite branch: the best reasoning stays expensive. Frontier intelligence is a premium meter; cheaper models handle the long tail. This is already the shape of the market, providers sell tiers, and the gap between the flagship model and the budget one is widening, not closing.

If that holds, optimization fragments by tier. The model a Fortune 500 buyer’s research agent uses is not the model a price-sensitive consumer uses, and they may surface different sources. SEO starts to mean: which model tiers cite us, and for which audiences?

What SEO becomes here:

• You optimize for model tiers, not just engines. Knowing whether your content is retrievable by the frontier models that high-value buyers use becomes a real competitive question.
• Authority compounds harder. Premium models lean on the strongest, most verifiable sources because their users are paying for accuracy. E-E-A-T stops being a ranking nicety and becomes the entry fee for premium-tier citation.
• The divide is a moat or a trap. Brands that earn frontier-model trust early get cited to the highest-intent audiences. Everyone else competes for the cheap-tier long tail.

Scenario C: Agentic commerce, when the meter buys things

The third scenario is the one moving fastest. Metered intelligence isn’t just answering questions; it’s acting. Google’s Universal Commerce Protocol (announced January 2026) and OpenAI’s Agentic Commerce Protocol are infrastructure for agents that research, recommend, and transact. At Google I/O 2026, agentic booking expanded to flights and hotels. AI-referred traffic to U.S. retail surged more than 800% year over year on Black Friday 2025.

When an agent spends its user’s metered budget to complete a purchase or booking, the “audience” you are optimizing for stops being human. SEO becomes agent readiness.

What SEO becomes here:

• Machine-readable proof beats persuasion. Agents don’t read your hero copy; they parse structured data, prices, availability, specs, policies, and reviews. Schema markup moves from optional to load-bearing.
• The cheapest tokens to parse win. An agent on a meter will favor sources it can extract cheaply and confidently. Clean structure is literally cheaper for the model, and that cost-efficiency becomes a ranking advantage.
• Conversion happens before the website. Like the declining click-to-call we’re already seeing in local search, the transaction increasingly resolves inside the conversation. Your job is to be the option the agent is confident enough to pick.

The no-regret moves: what wins in all three

Here is the useful part of any scenario exercise, find the moves that pay off no matter which future arrives. These do:

1. Become the most citable source on your topics, not the most ranked. Original data, first-hand experience, and clear answers are what every scenario rewards. This is the heart of generative engine optimization.
2. Make everything machine-readable. Structured data, clean hierarchy, direct answers under headings. It helps human SEO today and is mandatory for agents tomorrow, at no cost to either.
3. Build verifiable authority. Real authors, traceable credentials, primary sources. Abundance rations attention to trusted sources; premium tiers demand it; agents transact on it.
4. Shift your metrics now. Start tracking AI citations, brand mentions in answers, and branded-search lift alongside clicks. If clicks are the only number you watch, you’ll think you’re losing while you’re actually winning the new game.
5. Own the relationship the meter can’t. Email, community, brand. When the engine sits between you and the customer, the direct connection is the one asset no protocol can disintermediate.

My bet

I think the real future is a blend: abundance for everyday queries, premium tiers for high-stakes ones, and agents transacting across both. The through-line is that search stops paying for ranking and starts paying for trust the engine can verify and cheaply reuse. Altman’s “meter” isn’t only how you’ll pay for intelligence, it’s the lens for understanding why AI engines will get pickier about whose content they spend compute to surface.

The brands that win won’t be the ones who ranked highest. They’ll be the ones worth the tokens.

For the foundations underneath all of this, see our guides on GEO and AI Overviews SEO, and the related read on how Google replaced the search box with an AI agent.