What is dark social?

Dark social is sharing that happens where analytics cannot follow it: a link pasted into a DM, a screenshot dropped in a group chat, a recommendation given in a private Slack or Discord room. Alexis Madrigal named it in The Atlantic in 2012 after noticing that most social traffic arrived with no referrer and got filed as direct. The sharing is not hidden from people. It is hidden from measurement.

Why does dark social matter to a founder?

Because the most valuable sentence about your product is usually said there. "We use this, it is fine" in a private room moves a decision further than any ad, and it never appears in your analytics or your attribution model. Two things follow. First, your direct traffic number is not really direct, so a spike in it often means someone recommended you somewhere you cannot read. Second, your demand is always larger than what you can count, and reading the counted part as the whole picture makes you cut the channels doing the work.

How do you find the visible edge of dark social?

You cannot read the group chat, and you should not try. What you can do is watch the public conversations next to it. Private recommendations start as public questions surprisingly often, and the person who asked a room of friends will ask a subreddit a week later. The public edge includes what do you use threads, complaint posts naming a competitor, and migration announcements. Watching that edge is ordinary social listening. The other practical move is a tracked link per place you post, so a share out of one room is at least distinguishable from a share out of another. More on the whole path is in the dark social pipeline post.

Questions

Is dark social the same as direct traffic?

Not quite. Direct traffic is a bucket in an analytics tool for visits with no referrer. Dark social is the behavior that fills most of that bucket: links passed hand to hand in private. Some direct traffic really is typed URLs and bookmarks. The rest is a conversation you were not in.

Can I measure dark social at all?

Partly. Tracked links with a distinct code per place, a "how did you hear about us" field at signup, and a habit of asking new users where they saw you will recover a useful share. None of it is complete. Treat the number as a floor rather than a measurement.

Does AI search make dark social bigger?

It adds a similar blind spot. When someone asks an assistant which tool to use and acts on the answer, no click is recorded anywhere you can see. The habit that helps is the same one: watch where the answers come from, which is what share of model tracks, and show up in the public sources they draw on.