How to Get Found by AI Recruiters in 2026

For a long time your CV had two readers. A person, and a piece of software called an ATS that scanned for keywords before the person ever saw it. Every guide you have read is about pleasing those two. Match the keywords, keep it clean, get past the bot, land in front of the human.

There is a third reader now, and almost nobody is writing for it yet. It is an AI assistant, and it does not wait for you to apply. It goes looking. Someone asks their model “find me a React developer who is open to work,” and the model goes and reads real candidates and comes back with a shortlist. You are either in that shortlist or you are invisible to it.

I run FreeCV, and I can see this happening in our own numbers. We publish open, machine-readable CV feeds for people who want them. When I look at who actually reads those feeds, the single biggest visitor is not a recruiter or a person. It is ClaudeBot, Anthropic’s crawler. It is about 70 percent of the traffic. Add the other AI and search crawlers and roughly four in five reads are machines. The humans are the minority already.

So this is not a prediction. The AI reader is here, it is hungry, and most people’s CVs are completely unreadable to it. Let me explain why, and what to do about it.

Your CV now has three readers, and they want different things

It helps to be blunt about who you are writing for, because the three readers do not overlap as much as you would hope.

• The human recruiter spends six to ten seconds on a first pass. They want a clean layout, a clear title, and results they can scan.
• The ATS wants exact keywords that match the job description, in a format it can parse without choking.
• The AI agent wants structured data it can trust. Not a picture of a document. The actual fields: your title, your skills, your dates, whether you are open to work, in a shape it can read the same way every time.

Here is the trap. A PDF is built for the first reader and tolerated by the second. For the third it is a mess. When an AI model reads a PDF it is really looking at a flattened page and guessing where the columns were. Two of us tested this properly. We took one CV, rendered it into five file formats, and fed all five to four different AI models against an answer key. The thing that broke was not the fancy formatting. It was structure. The models dropped employers, merged roles, and lost fields, and the cheaper the model the worse it got. You can read that whole experiment in what each AI model lost.

What the AI reader actually wants: a machine-readable CV

A machine-readable CV is just your career as structured data instead of a laid-out page. Name, work history with real start and end dates, skills as a list, education, and a plain flag that says whether you are open to work. No columns to reconstruct, no fonts to see through, no guessing.

We built an open standard for exactly this called cv.json. It is free, anyone can generate it, and any tool can read it. When you publish a CV on FreeCV and claim your live link, you automatically get a cv.json feed at a stable public address like livelink.cv/yourname/cv.json. That address is the thing the AI reads. It is always current, because it points at your live CV, not a file you emailed six months ago and forgot.

This is the part people miss. A PDF is a snapshot. The moment you send it, it starts going stale. A live machine-readable feed is the opposite. You update your CV, the feed updates, and every AI that checks it sees today’s version. For a reader that comes back and re-reads on its own schedule, that difference is everything.

The living talent directory: where AI recruiters look

Reading one CV is useful. Searching thousands is where it gets interesting, and that is the piece we just shipped.

We turned every public, open-to-work CV on FreeCV into one searchable index and put a human face on it at freecv.org/people. Right now that is close to a thousand real people who have said they are open to work, searchable by skill, role and location, each one linking to their live portfolio and their machine-readable feed. We call it the living directory on purpose. It is not a pile of stale profiles. It ranks by relevance, puts people who are actively looking first, and favours recently updated CVs, because a directory of people who might have taken a job two years ago is worthless to anyone hiring today.

How an AI actually queries it

This is the part built for the machines, and it is worth understanding even if you never touch it, because it tells you what to optimize for.

We published an MCP server for cv.json. MCP is the standard that lets an AI assistant use outside tools, and ours gives any assistant three abilities: search the open-to-work index, fetch a person’s full machine-readable CV, and check that a CV is valid. Anyone can connect it. In Claude Desktop or any MCP client it is one line:

npx -y cvjson-mcp

Or connect over the hosted endpoint at https://freecv.org/api/mcp. Once it is connected, a recruiter can literally ask their assistant “find me nurses open to work in the Gulf” and the assistant runs the search, reads the matching CVs, and answers. No job board login, no scraping, no PDF parsing. It asks a question and gets structured people back.

You do not need to install any of that to benefit. You just need to be in the index it searches, with a CV clean enough that when the assistant fetches you, it understands you.

How to get found: five steps

Here is the actual playbook. None of it costs money.

1. Build a real CV. Start at the free builder. Fill in real dates, a clear job title, and your skills as separate items, not one blob. The AI reads those as fields, so give it clean fields.
2. Claim your live link. Publish your CV and claim your livelink.cv/yourname page. That is the moment your machine-readable cv.json feed goes live and becomes fetchable.
3. Mark yourself open to work. This is the flag that puts you in the directory an AI recruiter searches. If you never set it, you are publishing into the void.
4. Write a title an AI can match. “Registered Nurse” beats “Healthcare Professional.” The search matches whole words in your title, skills and location first, so put the words a recruiter would actually type where they carry the most weight. Our own guide on what to put on a resume covers this.
5. Keep it fresh. Update it every so often. The directory ranks recently updated profiles higher, and a live feed only helps you if it reflects who you are now.

Why this beats a job board for the AI era

A traditional job board or a big network is a walled garden. The data lives inside it, a login stands in front of it, and an AI cannot read it without permission it usually cannot get. Your profile there is only as useful as that platform lets it be.

An open machine-readable CV is the opposite. It is your data, at your own stable address, in a format anyone’s tool can read, with no wall in front of it. When the way people search shifts from “log into the platform and click around” to “ask my assistant and let it go read,” the open, machine-first CV is the one that gets read. The closed profile is the one that gets skipped.

I am not claiming networks are dead. I am saying there is a new front door, most people have not walked through it, and it is wide open and free. The people who show up in the machine-readable directory now, while it is early, are the ones the AI readers learn to trust first.

Frequently asked questions