If you want the conceptual grounding first — why X works as an acquisition channel and how it compares with Reddit on volume, intent and reply windows — start with X vs Reddit for B2B leads. This page assumes you are already convinced and want the operating procedure: what to watch, how to filter, when to reply, and what to do when the reply lands.
What changed, and why the channel is underworked
Two things happened at once. Reddit monitoring became crowded — there are now half a dozen tools competing to surface the same threads — while X monitoring quietly got harder, because the API tightened and the free tooling that used to cover it stopped working. The result is a channel with roughly the same volume of buying-intent conversation and a fraction of the competition.
The second change is behavioural. Founders, operators, and technical buyers increasingly post their evaluation process in public: the shortlist, the pricing objection, the migration pain. That is not a side effect of X being a social network. It is a function of the audience, and it is why a B2B tool can find qualified demand there that does not exist on LinkedIn, where the same person writes for their employer rather than themselves.
Build the signal set before you build anything else
Most X lead generation fails at the first step: people track their brand name and a category term, get a feed of noise, and conclude the channel does not work. The brand name is the least valuable thing you can track, because by the time someone names you they have already found you.
A working signal set has four layers. Competitor names catch people who are actively comparing. Pain phrases catch people who have the problem but not yet the category vocabulary. Workaround language — spreadsheet, manual, script, Zapier, cron job — catches the buyer at the exact moment their improvised solution breaks. Switching language — "migrating off", "cancelled our", "looking for an alternative to" — catches the highest-intent moment on the platform.
Write out fifteen to twenty phrases across those four layers before you configure a single alert. The layer that produces your best leads is usually not the one you expected, and you only find out by running all four.
Qualify in two passes
Raw keyword matching on X produces a brutal signal-to-noise ratio, worse than Reddit, because the same words appear in jokes, news commentary, and marketing copy. You need two filters, not one.
Pass one is mechanical. Exclude retweets. Exclude accounts with no bio and no profile image. Exclude posts that are themselves promotional. Set a minimum engagement floor only for discovery queries, never for intent queries — a buying-intent post with two likes is more valuable than a viral thread with two thousand.
Pass two is semantic. Somebody, or something, has to read the post and decide whether this person is describing a problem you solve, and how close they are to acting. This is the step that cannot be done with boolean logic, and it is the step that decides whether the channel is worth running. MentionSpot does this with an LLM and returns a score plus a stage; doing it by hand works too, and takes about fifteen minutes a day at low volume.
The timing rule
X has a far shorter useful window than Reddit. A Reddit thread keeps collecting readers from search for years; an X post is functionally dead within a day, and the replies that get read are the ones that arrived while the author was still looking at their notifications.
Practically: a reply inside the first hour has a high chance of being seen and answered. Inside six hours, moderate. After twenty-four hours you are writing for nobody. This is the single strongest argument for monitoring rather than periodic manual search — a daily check catches perhaps a fifth of your opportunities at a useful moment.
The corollary is that you should tune alerting to your actual working hours and accept that you will miss the overnight window. Trying to cover twenty-four hours with one person produces burnout and worse replies.
The reply workflow
Four steps, in order, every time.
Read the account, not just the post. Thirty seconds on their timeline tells you whether they are a real buyer, a competitor, or someone who posts the same complaint weekly for engagement.
Answer the question as if you had nothing to sell. If the useful answer is a competitor, say so. If the useful answer is "you do not need a tool for this yet", say that. This is not altruism; it is the only reply style that survives on a platform where everyone has seen a thousand pitches.
Mention what you built, once, plainly, with disclosure. One sentence. No link in the first reply unless they asked for one — links suppress reach and read as a pitch.
Move to DM only when invited. An unsolicited DM after a public reply is the fastest way to convert a warm prospect into a block.
A note on tooling honesty: MentionSpot finds and scores these posts. It does not write the reply, does not post on your behalf, and has no DM function. The judgement in the four steps above is the part that produces the conversation, and it has to be yours.
What to measure
Do not measure impressions. Measure four things: posts surfaced, posts you judged worth replying to, replies that got a response, and responses that became a conversation off-platform.
The ratio that tells you whether your signal set is working is the second one — surfaced to worth-replying. Below about one in ten, your keywords are too broad. Above one in three, they are too narrow and you are missing volume. The ratio that tells you whether your replies are working is the third, and if it sits below one in four the problem is almost always that you pitched too early.
| Metric | Healthy range | What a bad number usually means |
|---|---|---|
| Surfaced to worth-replying | 1 in 10 to 1 in 3 | Keyword set too broad or too narrow |
| Reply to response | Above 1 in 4 | Pitching before answering |
| Median time to reply | Under 2 hours | Checking manually instead of monitoring |
| Response to real conversation | Above 1 in 5 | Wrong intent stage, or wrong audience |
Where this breaks at scale
It breaks when you try to industrialise it. Five to fifteen genuinely considered replies a week from a real account with a real history is a sustainable rate that compounds. Fifty templated replies a week from a two-week-old account gets rate-limited, reported, and eventually suspended.
The other failure mode is delegating the writing to someone with no product knowledge. The reply only works because it contains something a marketer cannot fake — an actual answer to a technical question. If nobody on the team can write that answer in five minutes, the channel is not ready for you yet.
FREQUENTLY ASKED QUESTIONS
Questions about X lead generation
Does X lead generation still work now that reach is throttled?
Reach on original posts is unpredictable, but replies to a post appear in that author's notifications regardless of your follower count. Lead generation by reply is largely insulated from the reach problem, which is why it outperforms posting for most small accounts.
How many keywords should I track?
Fifteen to twenty phrases spread across competitor names, pain language, workaround language, and switching language. Fewer than ten usually means you are only catching people who already know your category exists.
Should I reply publicly or send a DM?
Publicly first, always. DMs from strangers are ignored or reported at a very high rate, and a public reply that is genuinely useful is also read by everyone else who finds the post later.
Does MentionSpot write or send the replies?
No. It monitors X and Reddit, scores each match for relevance and buying-intent stage, and gives you a ranked queue. It never posts, replies, votes, or sends messages on your behalf.
WORK THE CHANNEL NOBODY ELSE IS WORKING
Your competitors are all watching Reddit. X is still open.
MentionSpot monitors X and Reddit for buying-intent conversations in your category, scores every match, and hands you a ranked queue. You write the reply.
Get the 7-day pass