How to Find Leads on X (Twitter) in 2026

TL;DR

• X (Twitter) is full of real buying intent — people publicly tweet “anyone know a good [your job]?” The work is there; the trick is finding those tweets fast and replying like a helpful human, not a bot.
• The manual method is just good search plus good manners: build buying-intent queries with advanced search operators, check them daily, answer the question first, and only move to DMs when you’re invited.
• In 2026 the bot route is genuinely dangerous — X bans automated keyword replies and bulk DMs, and ban waves are hitting auto-reply tools. The safe edge is being early and being human. That is the one part of the job ClientRadar helps with, and it never auto-posts.

Why is X still a good place to find clients in 2026?

People treat X like a search bar for other humans. When someone needs a freelancer fast, they often just tweet the question: “Can anyone recommend a good web designer? Ours ghosted us.” “Looking for a photographer in Lisbon for next month.” “Need a copywriter who actually understands SaaS — recs?”

That is the highest-intent signal you can get for free: a stranger raising their hand in public, with a problem and a willingness to pay. Unlike Facebook groups, these tweets are wide open — you don’t need to be a member of anything to see them or reply. And unlike a cold list, nobody is annoyed that you showed up, because they literally asked.

The catch is the same one every platform has: the value is in catching the tweet while it’s fresh and replying like a person. By the time the tenth reply lands, they’ve usually already DM’d someone. So the whole skill is two things — finding the right tweets quickly, and answering well. Let’s do both by hand first.

How do I search X for buying intent?

X’s advanced search is the engine here, and it’s free. You can use the graphical form at x.com/search-advanced, but it’s worth learning a few operators so you can build precise queries and save them.

The core move is to combine intent phrases with your service and a few filters. A starting query for a web designer might be:

("looking for" OR "can anyone recommend" OR "need a") ("web designer" OR "website built") lang:en -filter:replies

Here’s what each piece does, and the operators worth knowing:

• Quoted phrases match exact wording — "looking for" is the language of intent.
• OR inside parentheses lets you stack synonyms for both the intent and the service.
• lang:en keeps results in your language (swap for es, de, etc.).
• -filter:replies hides replies so you see original asks, not noise.
• min_faves:5 filters out near-invisible tweets if you’re drowning in results (a web-search operator; it works in the search box even though the official API doesn’t expose it).
• since:2026-06-01 / until: bound the date range when you want only recent asks.
• near:"London" within:25mi is gold for local service providers.
• -filter:links can cut out spam and self-promo, since real asks rarely include a URL.

Build three or four of these for your exact service and the phrasings your buyers actually use. Watch for signal words like recommend, looking for, need, hiring, budget, quote, ASAP, and DM me — urgency plus a budget plus a direct ask is the strongest combination. (X’s advanced search operators are well-documented if you want the full list.)

Should I use the X API to automate this?

Short answer: almost certainly not, and in 2026 it’s harder and pricier than people expect.

On 6 February 2026, X replaced its tiered plans with pay-per-use as the default, and there’s no free tier for new developers — you pay roughly $0.005 per post read, capped at 2 million reads a month, with further price shifts in April 2026. For a solo service provider, standing up an API pipeline to read search results is a lot of cost and engineering for something your eyes can do in five minutes a day.

More importantly, the API tempts people toward the thing that gets them banned — automated replying. We’ll come back to that, but the honest takeaway is: for finding leads, manual search plus saved searches beats the API for almost everyone reading this. The API is for products, not for one person looking for clients.

How do I save searches so I don’t have to retype them?

Once a query works, don’t lose it. You have a few free options:

1. Save the search in X. Run the query, tap the ”…” menu on the results, and choose “Save search.” It then sits one tap away in the search box.
2. Bookmark the URL. A search like x.com/search?q=... is just a URL — bookmark your best three or four and open them as a morning routine.
3. Pin them as browser tabs you skim with your coffee and again after lunch.

Aim to check your saved searches once or twice a day. You’re not trying to live in the app — you’re trying to catch fresh asks within an hour or two, while you can still be one of the first useful replies.

What does a reply that actually wins look like?

This is the whole game, and it’s the same on every platform: the person didn’t tweet to get an advert. They tweeted a question. So answer the question.

A reply that earns the work usually has three parts:

1. Be genuinely helpful. Answer what they actually asked, even if part of the answer doesn’t involve hiring you. One real tip, one thing to watch out for. This earns the right to everything after it.
2. Show one proof point. Not your résumé — one relevant line. “Built a SaaS marketing site last month, happy to show you.” One true, relevant detail beats a paragraph of adjectives.
3. Make a soft ask. “Happy to send a couple of examples if useful” or “feel free to DM me and I’ll talk you through it.” You’re opening a door, not shoving them through it.

A worked example, replying to “Need a copywriter who gets SaaS — recs?”:

The thing I’d watch for is whether they can write for activation, not just the homepage — most SaaS copy dies in the onboarding emails. I write SaaS copy and just did a trial-to-paid sequence that lifted conversions for a client. Happy to share it if it’s useful — no pressure either way.

Notice it helps even if they hire someone else, proves relevance in one line, and the ask is soft. It reads like a person, because it is one. Keep it short; long replies read as pitches.

When is it OK to slide into the DMs?

Let the conversation earn the DM. If they reply warmly, or literally say “DM me,” continue there. That’s invited, and it’s fine.

What is not fine — and what gets accounts suspended in 2026 — is cold, bulk, or automated DMs. X’s rules are explicit: you may not send unsolicited direct messages in a bulk or automated manner, and you may only DM someone automatically if they’ve clearly indicated they want to be contacted. (See X’s automation rules.) The safe rule of thumb: a human, replying in public, then moving to DMs by invitation. Never a tool firing messages at a keyword list.

Why are bots and auto-reply tools so risky right now?

This is where a lot of “X lead gen” advice will get you burned. X’s enforcement against automation got dramatically stricter in 2026, with ban waves hitting accounts that use unauthorized auto-reply and auto-DM tools. The platform’s own policy spells it out: sending automated replies to posts based on keyword searches alone is not permitted, and bulk unsolicited mentions on broad topics are prohibited.

That’s the exact thing many “reply automation” products do. If you’ve looked at ReplyGuy-style auto-reply tools, this is the core risk — automating the social part is the line you don’t want to cross, because the downside is losing the account you spent years building. A genuinely useful comparison:

Approach	Speed	Risk to your account	Feels like
Manual search + reply	Slower	Lowest	A helpful human
Free keyword alerts (F5Bot-style)	Fast alerts, manual reply	Low	You, just faster
Auto-reply / auto-DM bots	Fastest	High (ban waves)	A bot, because it is one

The middle row is the sweet spot for most people: get alerted fast, but keep the human on every reply.

Where does ClientRadar actually help — and where doesn’t it?

Here’s the honest version. Everything above, you can do by hand, and for your first weeks you should, because it teaches you what a real buyer sounds like on X. The wall you’ll hit is the same one freelancers hit everywhere: keeping good searches running across X (and Reddit, Facebook, LinkedIn), all day, and being early and warm every time. Nobody has that kind of time while also doing the actual work.

That’s the gap we built ClientRadar to fill — and only that gap. It watches the searches you care about, flags the tweets that show real buying intent, scores that intent 0 to 100 with the reason why, and drafts a reply in your own voice for you to read, edit, and send yourself. Your leads and notes stay local in your browser as a simple CRM.

The line we will not cross is the one above: ClientRadar never auto-posts and never auto-DMs. There’s a human — you — on every single send. That’s not a limitation we’re apologizing for; in 2026 it’s the entire point. The tools that auto-post are the ones getting suspended.

If finding a few real buyers a week sounds more useful than chasing hundreds of mentions, that’s the trade we’re built for. One signed client usually covers the year.

Your quick checklist

• Build three or four buying-intent queries with advanced search operators for your exact service.
• Save them in X, or bookmark the search URLs, and check them once or twice a day.
• Skip the API and skip the bots — for one person finding clients, neither is worth it.
• Reply in public: answer first, one proof point, soft ask. Keep it short.
• Move to DMs only when invited. Never cold, bulk, or automated.
• One gentle follow-up, then let it go. Track every conversation so nothing slips.

You can do every step here by hand, and that’s the right way to start. When running searches across platforms all day becomes the bottleneck, that’s the moment a tool earns its place. You can try ClientRadar free (no card needed) and see what it flags in the searches you already care about. Either way, the asks are out there — go be the helpful reply.