Reddit intent detection that tells buyers from browsers

Reddit intent detection reads a post for signs its author is about to choose a product, then scores how close they are to buying. Reddix scores every post it finds from 0 to 100 against your product, so the people asking for a tool like yours come first.

A keyword alert can’t tell these apart

All four posts mention “CRM”. A keyword alert sends you every one of them. Intent detection ranks them by how close the writer is to choosing a tool.

  • low (8): “Our CRM migration finally finished. Here’s what I learned after three months of it.” A finished project. There’s nothing left to choose.
  • research (31): “How do small teams usually keep track of deals? Spreadsheets, or is a CRM worth it?” An early, open question. Worth a helpful answer, not a pitch.
  • warm (58): “Our sales notes live in five different places. Is a CRM overkill for a 4-person agency?” A clear problem a CRM solves, and the writer is weighing it.
  • hot (86): “Looking for an alternative to Pipewell, budget under $30/month. What do you all use?” A direct ask, a tool they want to leave, and a budget.

Illustrative posts and scores.

From search to saved lead, in four passes

Each scan runs the same four passes. The cheap ones go first, so the AI only reads posts worth reading.

  1. Search like a buyer. From your product profile, the AI writes the phrases a customer types when they have the problem you solve, and Reddix searches Reddit with them. (6–8 buyer phrases; Up to 7 searches per scan)
  2. Throw out the obvious noise. Rules run before any AI: posts already scored this week, posts too short to judge, launches, hiring threads, giveaways and AMAs. A buying phrase or a competitor’s name rescues a post from those rules. Posts that name a competitor or ask directly go to the front. (7-day memory; Up to 30 posts go on)
  3. Score against your product. The AI reads each post next to your product description, features and competitors, and returns a score from 0 to 100, a level, a one-line reason and how confident it is. (Batches of 10; Score, reason, confidence)
  4. Add what the words prove. Some phrases are harder evidence than any model’s guess, so they add points after the AI. Posts that finish at 35 or above are saved as leads, best first. (Saved at 35+; Top 25 per scan)

What each score band means, and what to do

The AI scores against this rubric. The level you see on a lead is set by the final score, after the signal points are added.

LevelScoreUsuallyWhat to do
Hot70–100Asks for a tool, names a product they want to leave, or talks budget.Reply today, while the thread is still active.
Warm40–69Describes a problem you solve, or is comparing options.Answer the problem first. Mention your product if it fits.
Research25–39Exploring, curious, not choosing yet.Reddix keeps the ones at 35 and up. Help if you can add something.
Low0–24Off-topic, already solved, self-promotion or news.Not saved, and not scored again for 7 days.

Signals that add points after the AI

  • +25 — A competitor you named: “moving off Pipewell”
  • +15 — Pricing language: “budget”, “per month”, “how much”, “cheaper”
  • +10 — “Alternative to”: “an alternative to Pipewell”

Points are added to the AI’s score and capped at 100.

Paste a post and watch the first pass

This runs the same rules the scanner uses, in your browser. The AI score comes from Reddix during a scan; here you set it with the slider to see how the signals move the result.

What intent detection doesn’t do

  • It runs when you start a scan. Reddix doesn’t watch Reddit in the background. Each scan is a fresh look at the subreddits you chose, at most once every 5 minutes.
  • It reads public posts only. It sees what Reddit search can reach. Private communities, removed posts and messages are out of reach.
  • It never posts for you. Reddix finds the thread and drafts a reply. You read it, edit it and post it from your own account.
  • It can be wrong. Reddix saves the AI’s one-line reason and confidence with every lead, and shows the reason on the card. When the AI was less than 80% sure, Analyze gives the lead a second read, with its comments, before you spend time on it.
  • It’s only as sharp as your profile. The AI scores against your description, features and competitors. A vague profile gets vaguer scores. You can sharpen all of them in Settings.

Intent detection, answered

What is Reddit intent detection?
It’s reading a Reddit post for signs its author is about to choose a product: asking for recommendations, comparing options, naming a tool they want to leave, or mentioning a budget. Reddix scores each post from 0 to 100 on how likely it is to become a customer for your product.
How is it different from keyword alerts?
A keyword alert fires on any post that contains the word, so a launch announcement and a buyer asking for help look the same. Intent detection reads what the post is asking for and scores it against your product, so the buyer comes first and the announcement is skipped.
Which phrases count as buying signals?
Requests and comparisons such as “recommend”, “looking for”, “alternative to”, “switching from” and “what do you use” send a post to the AI first. Three signals also add points after the AI scores it: a competitor you named (+25), pricing words such as budget or per month (+15), and “alternative to” (+10).
Why did a post score lower than I expected?
Open the lead: it shows the AI’s one-line reason and how the score adds up, and when the AI was unsure, Analyze gives it a second read. The usual causes are a post that’s already solved, one that asks about a different problem, or a thin product profile. A clearer description, features and competitors in Settings give the AI more to match against.
Does Reddix scan Reddit automatically?
No. A scan runs when you press Scan now, on Home or in the Scanner, at most once every 5 minutes, and looks at the subreddits you chose. Every post it saves lands in the Scanner as a new lead, with its score and level.

Keep going