Your next signup is an AI agent. Is your marketing ready for it?

TL;DR: The PLG-versus-sales-led debate is stale. The live question is that a growing share of product discovery, evaluation, and soon purchasing runs through AI agents, not people. That means marketing to two buyers at once: a human who reads your homepage, and a machine that reads your docs, your pricing, and your llms.txt.

On July 11, Netlify launched netlify.ai, a version of their website built for AI agents instead of humans. Not a docs page. A website, with onboarding and contextual guidance, for software. Their CEO’s framing: “autonomous agents are going to be a whole new user base.”

Sit with that for a second, because it is a marketing decision, not an engineering one. A company looked at its funnel and concluded the next cohort of visitors cannot see the hero image, will not watch the demo video, and does not care about the customer logos. So they built a second front door.

Meanwhile, most of my feed is still arguing about whether PLG is dead. Wrong question. PLG is splitting. The motion where a human wanders in, tries the product, and swipes a card is being joined by a motion where an agent does the wandering, the trying, and increasingly the recommending. If your marketing only serves the first motion, you are optimizing a shrinking surface.

The buyer with no eyes

I run digital and performance marketing for a CPaaS and CCaaS company, which means a large part of my audience has always been developers who evaluate through docs and APIs before they ever talk to sales. What is happening now is that this evaluation style is escaping the developer niche and becoming the default, because the evaluator is not a person anymore.

Netlify coined a name for this in January 2025: agent experience, or AX - the experience an AI agent has as the user of a product or platform. How well an agent can discover your service, call it, and recover when something breaks now shapes whether your product gets used at all. They even ship an open-source scorer for it.

You can dismiss this as one infrastructure company talking its book. The pricing research says otherwise. Ibbaka’s 2026 B2B SaaS pricing predictions put it bluntly: “Buyer agents will screen you in or out before a human ever hits your website, making opaque pricing effectively invisible.” Invisible. Not unpersuasive, not weak. Absent from the consideration set, because the machine doing the shortlist could not parse a “Contact us” button.

And the money is moving to match. Deloitte’s 2026 technology predictions cite Gartner’s forecast that by 2030 at least 40 percent of enterprise SaaS spend shifts toward usage-, agent-, or outcome-based pricing. When the unit of purchase becomes “what an agent did,” the unit of marketing becomes “what an agent could find out.”

The machine doing the shortlist could not parse a “Contact us” button. That is not a weak impression. It is no impression.

Two buyers, one budget

Here is the reframe I would take into a 2026 planning cycle: every marketing surface now has two readers.

The human reader is who we have always served. Narrative, proof, pricing psychology, brand. None of that goes away, because a human still signs the contract and a human still gets fired for a bad purchase.

The machine reader arrives earlier in the journey, with no patience and no peripheral vision. It does not infer. It parses. It is assembling a comparison table for its human, and your job is to be a complete, unambiguous row in that table.

Honesty requires a caveat: today, most agent involvement is routing, not purchasing. A person asks ChatGPT or Claude what to use, the model reads whatever it can reach, and a human lands on your site with the shortlist already formed. The fully autonomous buyer with a corporate card is still mostly a demo. But the routing step is already deciding who gets the visit, and the transaction layer is being built in public right now. You do not wait for the bridge to finish before buying land on the far side.

What marketing to agents concretely means

This is the part I wish more of the discourse would get to. Five surfaces, in the order I would fix them.

1. Machine-readable pricing. If an agent cannot retrieve your pricing, you lose the comparison by forfeit. Publish real numbers, or at minimum a structured description of how pricing works and what drives it. The decade-old B2B instinct that hiding pricing creates sales conversations now creates exclusion instead.

2. An llms.txt and markdown mirrors. llms.txt is Jeremy Howard’s proposal for a curated, plain-markdown summary of your site at a fixed path, designed for context windows that cannot swallow your whole website. It takes an afternoon. This site has one, and I work on answer-engine visibility in my day job, so I have watched what happens to content that models can actually read versus content buried in JavaScript: one of them gets cited.

3. Docs as the landing page. Agents read documentation the way humans read homepages. If your docs are gated, stale, or thin, that is your brand to the machine reader. The classic PLG playbook said developers judge you by your docs. The agent era generalizes it: every buyer’s agent judges you by your docs, even when the buyer is a CFO.

4. Crawl access as a channel decision. Check what your CDN and robots.txt actually serve to AI crawlers, not what you think they serve. Plenty of sites block the very models their buyers ask for recommendations. That is a media budget of zero on the fastest-growing referral surface.

5. Attribution that expects machines. An agent does not click a CTA, fill a form, or accept a cookie banner. It reads, summarizes, and sends a human later, often direct or brand-search. If your dashboards treat that as “dark traffic,” your best channel will look like nothing is happening. Start tagging AI referrers and watching branded direct alongside them.

The operator takeaway

Nobody on your leadership team owns the machine reader yet. Product thinks it is marketing’s problem because it looks like traffic. Marketing thinks it is product’s problem because it smells like APIs. That gap is exactly where the next distribution advantage sits, and it is claimable this quarter by whoever writes the first internal memo.

I think that should be the marketer. We are supposed to be the people who understand the buyer. The buyer just grew a second pair of eyes that read markdown, and it is already outside, deciding whether your site makes the shortlist.