Boolean and X-ray search, after the ground moved.

Two things happened to sourcing in the last few years, and they pull in opposite directions. LinkedIn closed most of its public data to search engines, which made the single most-taught sourcing technique far less useful. At the same time, language models became good enough at writing search strings that the syntax knowledge those techniques depended on stopped being scarce.

The result is a lot of advice that is confidently out of date. What follows is what still holds.

What actually broke on LinkedIn

X-ray search meant using an external search engine to look inside a site rather than using that site's own search. For years the canonical example was:

site:linkedin.com/in "Director of Engineering" "Berlin"

That worked because LinkedIn let search engines index the useful parts of a profile - the headline, the location, the current title. It does not any more. LinkedIn stripped that metadata from what crawlers can see and put detailed profile information behind a login.

The important nuance, and the part most guides get wrong: X-ray is notdead. Profile URLs are still indexed. Searching a person by name still confirms they exist and finds their profile. What you have lost is the ability to filter - to ask Google for everyone with a given title in a given city and get a usable list back. That specific capability is gone, and no amount of clever syntax brings it back.

What replaced it

The most direct substitute exploits something LinkedIn does still want indexed: its advice articles. Those pages carry named contributors alongside their job titles, and they rank well.

site:linkedin.com/advice/ intext:"Director of Engineering"

This is genuinely useful, and it is also fragile. It works because of a decision LinkedIn made about a content programme, not because of anything structural. Treat it as a technique with a shelf life.

The more durable move is to stop treating LinkedIn as the index and start searching where your candidates actually leave evidence of their work: code hosts, personal sites, conference programmes, professional bodies, published papers, and public documents. Those sources were always better signal. LinkedIn's convenience simply made them easy to ignore.

The operators that still matter

Eight operators cover almost everything worth doing. The rest are variations you can look up when a specific search needs them.

Two practical notes that cause more failed searches than anything else.OR must be capitalised or Google treats it as a word. And excluding noise matters more than including keywords: a search for developers returns mostly job adverts until you strip them out with-jobs -hiring -apply -careers.

Exposed directories, and where the line is

One technique deserves separate treatment because it sits closest to the ethical line. Web servers configured without an index page will list their contents, and those listings get indexed:

intitle:index.of (resume OR cv) filetype:pdf

This surfaces documents nobody deliberately published. That is exactly why it works, and exactly why it deserves care. Something being publicly accessible does not mean it was intended for unrestricted use, and a CV sitting in a misconfigured folder was not put there as an application to you.

The test I would apply: could you tell this person how you found them without embarrassment? If the honest answer is no, the problem is the method, not the phrasing. Under GDPR you also need a lawful basis to process what you collect, and you need to be able to answer that question if they ask it.

What AI changed, and what it did not

A junior recruiter on my team recently generated a twenty-eight term Boolean string with an AI assistant. It was better than what I would have written by hand, and it took her seconds against the fifteen or twenty minutes it would have cost me.

I want to be honest about what that means rather than defensive about it. The syntax was a gatekeeping skill. It kept out capable recruiters who did not have the patience for punctuation, and it rewarded memorisation over judgement. Losing it is not a tragedy.

What did not transfer to the machine is the part that was always harder. A model will write you an excellent string for the wrong search. It does not know that this market calls the role something else, that the title you were given by the hiring manager does not exist at the companies you should be targeting, or that the first page of results is full of people who look right and are not. Deciding what to search for, and judging what comes back, is now the whole job.

Which makes the practical advice straightforward: let the model write the string, and spend the time you save on the intake conversation and on reading results properly.

Common questions

Does LinkedIn X-ray search still work in 2026?

Partly. LinkedIn removed job titles, locations, and most profile metadata from what search engines can index, so the classic site:linkedin.com/in pattern no longer returns the structured detail it used to. Profile URLs are still indexed, so you can still confirm that a person exists and find their profile by name. What you cannot reliably do any more is filter by title or location from Google alone.

What replaced LinkedIn X-ray search?

The most useful current substitute is searching LinkedIn's advice articles, which are heavily indexed and carry named contributors along with their job titles: site:linkedin.com/advice/ intext:"Director of Engineering". Beyond LinkedIn, the more durable answer is to search the places your candidates actually publish - GitHub, personal sites, conference programmes, professional associations, and published documents.

Is Boolean search still a skill worth learning?

The syntax is no longer the scarce part. A language model will write a twenty-term string in seconds, and it will usually be a good one. What it cannot do is decide what you should have been searching for. Knowing which terms a specific talent market actually uses, and being able to tell a promising result from a plausible one, is what still separates a good search from a busy one.

Is sourcing candidates this way legal and ethical?

Searching public information is legal in most jurisdictions, but public and fair game are not the same thing. Under GDPR you still need a lawful basis to process personal data, and you should be able to explain where you found someone if they ask. The practical test: if the person would be surprised or uncomfortable to learn how you found them, reconsider the method rather than the wording.