What 'Agent-Ready' Means When You're Not Selling to Agents Yet

Most operators I talk to dismiss agent-readiness as premature. "I do not sell to AI agents," they say. "I sell to humans. This is a problem for next year."

I understand the position. I also think it is wrong, and I want to explain why, without the futurist hand-waving.

When I made my own studio site agent-ready last month, I did it as an experiment. The site went from scoring 23 out of 100 on a public agent-readiness audit to passing 13 of 13 applicable checks. None of my clients are AI agents. None of my leads come from AI agents yet. So why did I spend the time?

Four reasons. Plus three audits you can run on your own site this afternoon to see whether you are already losing visibility you do not realise you have.

Reason 1: Search has already shifted, and you did not get the memo

The thing I notice most when I talk to operators is that they still imagine search as ten blue links. It is not.

When I search for "best self-hosted CRM" in the autumn of 2026, I do not get ten blue links. I get an AI-generated answer at the top, citing three sources. If your site is one of those three sources, you get traffic. If it is not, you get nothing.

The question every operator should be asking is: what makes a site the kind of source an LLM cites?

The answer involves a lot of things, but a non-trivial fraction of them are structural. Clear schema markup. Stable URLs. Content that is identifiable, attributable, and machine-parseable. Sites that are agent-ready tend to also be cite-ready. The two skill sets overlap by 70 to 80 percent.

You are not preparing to sell to agents. You are preparing to remain visible in a search environment where the first reader is always a machine.

Reason 2: Agent-discovery economics are starting to favour first movers

In every previous wave of new discovery surfaces, the operators who built for the surface early enough captured disproportionate share when the surface became mainstream.

I watched it happen with SEO in the late 2000s. I watched it happen with mobile sites in the early 2010s. I watched it happen with structured data and rich snippets a few years after that. The pattern is identical: the surface exists, most people ignore it, a small number invest early, the surface goes mainstream, and the early investors hold the high-traffic positions for years.

Agent-discovery is on the same arc, and I would argue it is earlier than any of the previous waves were when smart operators started moving. The cost of being early is one weekend of work on your site. The cost of being late is a multi-year traffic deficit you spend money trying to close.

I am not betting that agents are the future of all web traffic. I am betting that they will represent enough of it, soon enough, that a one-weekend investment pays back many times over. The expected value maths is straightforward.

Reason 3: Future-proofing for things you cannot predict yet

I run my business on the assumption that I do not know what the next five years look like, and I should build everything as if multiple futures are possible.

Agent-readiness is a low-cost option on a specific future. If agents become a major traffic source, I have built the right plumbing. If they do not, I have built better-structured content that performs slightly better in conventional search anyway. The downside case is "I made my site marginally better." The upside case is "I captured a discovery surface that became significant before my competitors noticed."

I take that trade every time. The only mistake is paying a lot to make it. Done correctly, the work is maybe 8 to 12 hours total for a small site.

Reason 4: It is genuinely cheap to do once you know the layers

The cheapest version of agent-ready is small. The expensive version is enormous. Most operators conflate the two and assume they need to build the expensive version. They do not.

Here is the cheap version, which is what I started with:

• A clear robots policy that allows the agents you want and explains the relationship
• A small number of well-formed metadata endpoints that describe what the site is, who runs it, and what services it offers
• Structured content with stable identifiers, so a machine reading the site can reliably tell what each page is and what it relates to
• A small number of named "skills" or "capabilities" that describe what the site can be asked to do

That is it. None of those things require new infrastructure. None require new code beyond a small theme update. None cost money. If you are running on WordPress with reasonable hosting, the entire stack can be added in an afternoon.

The expensive version, the full machine-readable storefront with a transactional agent endpoint, is a different conversation. You do not need it yet. The cheap version is what you start with, and it captures most of the value.

Three audits you can run yourself

If you want to know where your site sits today, here are three quick checks I run on any site I am evaluating.

Audit 1: the public scorer. A public service exists that scans any URL and rates it from 1 to 100 on agent-readiness. Go and run your own site through it. Most operator sites score in the 20s. Mine scored 23 before I did anything. Knowing your starting number gives you a benchmark to improve against.

Audit 2: the curl test. Open a terminal, fetch your home page using curl with a generic agent identifier, and read what comes back. Can you, as a human reading raw HTML, tell what the site is and what it offers? If you cannot, no machine will be able to either. Most pages fail this test because the content is fragmented across hidden JavaScript components.

Audit 3: the LLM brief. Open ChatGPT or Claude. Paste in your site URL and ask: "Tell me what this company does, who they sell to, and what they would offer me if I wanted to hire them." The quality of the answer is a leading indicator of how the same model would represent you in a real search response. If the answer is vague, generic, or wrong, the model does not understand your site, and neither will any agent built on top of it.

Running those three audits takes 20 minutes. The output is a punch list of the things to fix, prioritised by what an actual machine reader is struggling with.

The playbook

Here is what I recommend if you take any of this seriously.

1. Run the three audits above this week. Write down the starting score.
2. Pick the lowest-hanging fix from audit 3 and ship it. Usually it is rewriting your home page to lead with what you do, not who you are.
3. Add the cheap version of the agent-readiness layer over a single weekend.
4. Re-run audit 1 a fortnight later and confirm the score moved.
5. Decide based on what happens next whether to invest in the expensive version.

This is not a future thing. The discovery surface is live now. The operators winning on it are running the cheap version. You can be one of them.

For the technical detail of how I built my own layer, my cornerstone on the agent-readiness build walks through every part. For the infrastructure thinking that sits underneath it, the four hooks I use to evaluate any tool covers the wider stack. More about how I work is on the about page.

For the full agency-grade build, Weir Digital Media's complete guide to agent-ready websites is the longest-form treatment available.

Built by Weir Digital Media.