Blog · Reference

Every X search operator that matters for finding buyers

19 August 2026 · 10 min read

X search is far more capable than the box on the homepage suggests. Almost none of it is documented in one place, and most guides list the operators without saying which ones actually change your results. This is the reference, plus the part that matters: which combinations put a person in front of you instead of a wall of vendors.

Everything here works in the search box on x.com. No API, no account beyond a normal one.

Content operators

Two of these deserve more than a row. url: is the most underused operator on the platform: url:competitor.com returns everyone sharing, praising or complaining about a specific product, without you having to guess how they spell its name. And lang: matters more than people expect — a query without it mixes markets and buries the language you can actually reply in.

People operators

to:competitor is the sharpest people operator for this job. Everyone publicly complaining at a company is, by definition, a customer of theirs with a problem — and the ones who get no answer are the ones worth talking to.

Time operators

The date window is the single most valuable operator for buying intent, and the one most people skip. A query with no window returns conversations that closed months ago, which is the fastest way to waste an hour and conclude that the channel does not work.

Rules of thumb: seven days on a broad category, thirty on a narrow one. If a seven-day window on a broad term returns nothing, the query is wrong before the market is.

Filters

-filter:links is the highest-yield filter in this whole table and almost nobody uses it. Vendors, agencies, newsletters and affiliates nearly always attach a URL. People describing a problem almost never do. One operator, most of the marketing gone.

Engagement thresholds

Use these gently or not at all. A small floor like min_faves:1 removes pure bot output. Anything higher removes exactly the people you want: someone quietly asking their two hundred followers for a recommendation gets almost no engagement, and they are the best lead of the week. High engagement thresholds surface popular posts, and popular is not the same as buying.

There is one good use for min_replies:: finding threads where a category is being debated. Set it around five and you get discussions rather than individual asks — useful when you want to understand how people talk about your market before you go looking for buyers in it.

The four combinations that work

Operators on their own are trivia. These four shapes are what actually returns people. Replace the bracketed parts.

1. The recommendation ask

("any recs for" OR "any recommendations" OR "looking for a" OR "what do you use for") [category] -filter:retweets -filter:links lang:en since:2026-08-12

The -filter:links is doing most of the work here. Without it, half the results are vendors answering someone else.

2. The switcher

("moving off" OR "migrating from" OR "switching from" OR "leaving" OR "cancelling") ([competitor] OR [competitor]) -filter:retweets lang:en since:2026-08-01

Keep links in for this one — switchers often link the pricing page that made them leave.

3. The complaint at a competitor

to:[competitor] (broken OR "still waiting" OR "no reply" OR unusable OR "so slow") -filter:retweets lang:en since:2026-08-05

This one is unusual and very effective. It searches replies aimed at a company, which is where unhappy customers go first.

4. The stated constraint

[category] ("budget is" OR "under $" OR "before the end of" OR "by next month" OR "we need this by") -filter:retweets lang:en since:2026-07-20

Rare and worth every minute. A number or a date in a post means somebody has already decided to spend.

Our intent query builder assembles these from your offer if you would rather not write them out, and the intent score checker will tell you whether a result you found is actually a buyer.

Operators that look clever and are not

• filter:verified — verification says somebody pays for a subscription, not that they buy software. It mostly filters out your buyers.
• Hashtags — people in trouble do not tag their trouble. Hashtag searches return marketers talking to marketers.
• High min_faves — optimises for popularity, which is inversely correlated with someone quietly asking for help.
• near: and within: — geo data is sparse on X and getting sparser. For most B2B software it removes far more real results than noise.
• Very long OR chains — past roughly eight alternatives a query gets slow, unpredictable, and impossible to debug when it stops working.

Why your query rots

A query that worked in March returns nothing useful in August, and it is rarely because the market moved. Three things happen.

Vocabulary drifts. A new competitor appears, a product gets renamed, a phrase falls out of use. Your named alternatives go stale faster than anything else in the query.

Noise accumulates. A new spam pattern latches onto your keywords, and unless you keep adding exclusions, the signal-to-noise gets worse every month. A query that is not getting more precise is getting worse.

The operators themselves change. X has quietly altered filter behaviour more than once. Anything undocumented should be treated as working until it does not — test rather than trusting a guide, including this one.

The practical answer is to keep your queries somewhere you will actually revisit, add an exclusion every time you see the same junk twice, and re-read the whole set once a quarter. That is the maintenance nobody mentions when they publish an operator list.

Where this fits

Operators find candidates. They do not tell you which candidate is a buyer, and they do not write the reply — those are the two steps that decide whether this channel produces customers. The guide to finding buyers on X covers the whole method, a mention is not a buyer covers the triage, and replying without sounding like spam covers the last step.

Or skip the query maintenance entirely: run it on your offer and Askline writes and maintains the searches for you.