How to Stay Visible in the Age of AI Search

“SEO isn’t going to be a thing anymore,” said Aravind Srinivas, CEO of Perplexity. Apple testified that Safari searches are falling — in fact, declining for the first time in 22 years — wiping $150 billion off Alphabet’s value in a single day. “People are using AI,” said Apple’s senior VP Eddy Cue during the DOJ’s antitrust trial against Google, calling it a fundamental shift eroding traditional search. Google insists everything is fine. But behind the spin, the ground is clearly shifting.

What’s under threat isn’t just SEO, but the foundation of Google’s business model: a monopoly over online discovery that fuels a $300B ad machine, now targeted by two U.S. antitrust lawsuits. The most lucrative piece of it? A $20 billion-a-year deal to be the default search engine on Safari — worth around 36% of Google’s mobile search ad revenue. But even that grip is loosening. Apple has confirmed it’s actively considering adding AI-powered search providers like OpenAI and Perplexity to Safari in wake of declining search numbers. If that default changes, the dominoes fall.

In a short statement, Google claimed “overall query growth” is still rising, including from Apple devices — while carefully avoiding any mention of Safari’s reported decline. The language was vague (“devices and platforms”) and offered no real data, just the usual nod to “new ways to search” like voice and Lens. Which is convenient. Because when your business depends on appearing indispensable, query growth starts to look a lot like a vanity metric — especially if people are just searching more because they’re finding less.

And while Google insists nothing has changed, things are clearly changing. In April, OpenAI reported over 1 billion weekly web searches through ChatGPT’s search feature. By February, it had reported more than 400 million weekly active users. The shift isn’t speculative. It’s happening in real time — and we will at some point reach a critical mass where it simply doesn’t make sense to use a legacy search engine for answers that AI systems can generate faster, more directly, and with better context.

For 25 years, search meant Google. Get crawled, get ranked, get paid. SEO was the game, and we all played it. But now, a new paradigm has emerged — one that doesn’t crawl and index, but reads and reasons. ChatGPT, Perplexity, Claude, and others don’t show links. They show answers (albeit with citations). Answers built from a soup of scraped content, weighted by relevance, clarity, and reputation — not page authority. The web used to be about ranking. Now it’s about reasoning. If your content isn’t part of the synthesis, it isn’t seen.

But how will anyone stay visible in this new world of online? It’s the golden question, and most people still don’t have an answer. But there’s one new, lightweight experiment that could offer serious upside: llms.txt.

It’s a brand-new standard. Think of it as robots.txt for the LLM era, a simple file that tells AI models what your site is, how it’s structured, and what’s worth surfacing. It doesn’t game the system. It guides it.

Already being adopted by early movers like Cloudflare, Perplexity, Anthropic, and ElevenLabs, llms.txt is quickly becoming the gateway to visibility in a world where models generate answers — not search results.

Whether it works as intended is still an open question. But as always with AI, just as one system lands, another disruption — with a capital D — is already coming into view.

Agents.

If LLMs changed how people search, agents will change who searches altogether. No more users typing queries. Just AI agents browsing, comparing, retrieving, and acting — on your behalf, or theirs.

Google’s already preparing. The release of its Agent2Agent protocol last month is laying the groundwork for how AI agents will talk to each other, transact, and make decisions across different models and architectures.

Which raises the next existential marketing question: how do you advertise to an agent?

They don’t browse. They don’t click. They don’t care about your brand voice. They care about structured data, machine-readable value, and trusted logic. Visibility in this new era won’t come from ranking high — it will come from being useful, parsable, and executable.

Today, that might mean llms.txt. Tomorrow, it might mean surfacing through a real-time agent marketplace, or being whitelisted in an enterprise knowledge graph.

In the meantime, ask yourself this: if your content had to convince a machine — no style, no story, just facts — would it pass?

Because when agents start making decisions, the answer is the ad. And you only get picked if you make sense.

That said, ads find a way. They always do. Google will build its agent layer. OpenAI will likely monetise its answers at some point. Perplexity already includes sponsored links. But the shape of influence is shifting — away from eyeballs and toward inference. You won’t be bidding for impressions. You’ll be bidding for inclusion in a reasoning chain.

The decline in Safari searches might signal a shift in behaviour, but let’s not get carried away. Google still holds the default slot on Apple devices — a $20 billion deal that Apple hasn’t ditched yet, and one Cue admits keeps him up at night. Perplexity and ChatGPT may dominate the narrative, but Google still owns the numbers: 5 trillion searches a year, and AI Overviews now seen by 1.5 billion users a month. The AI tide is rising — but this isn’t (yet) an obituary. It’s a diagnosis of decline.

Follow the lines…