> For the complete documentation index, see [llms.txt](https://themochi-app.gitbook.io/documentation/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://themochi-app.gitbook.io/documentation/module-4-ai-setup-and-intelligence/ai-tags.md).

# AI Tags

## AI Tags: Auto-Tag Leads by Intent

**What it is** — An AI tag is a label that Mochi applies to a lead automatically, based on a plain-language prompt you write instead of a fixed keyword. Mochi reads the conversation, checks it against your prompt, and applies (or removes) the tag as the conversation evolves — catching intent even when a lead phrases it ten different ways.

**Who can use it** — Only **Creator and Manager** can create or edit tags, write AI prompts for them, or run a Retag. Other roles can see and filter by tags on leads they already have access to. Available on **web and iOS**, both at `Settings > Tags`.

**Before you start** — A lead needs some conversation history before an AI tag has anything to evaluate; a brand-new, empty conversation won't match any prompt yet.

### How AI tagging works

Once a tag has an AI prompt attached, Mochi:

1. Reads the full conversation for each lead.
2. Checks it against the tag's prompt, matching on **intent**, not just exact wording. A "Hot Lead" prompt built around urgency catches "I'm ready," "where do I sign up," and "how do I get access" as the same signal.
3. Re-evaluates automatically every time a new message comes in, so a tag can start — or stop — applying as the conversation moves.
4. Optionally requires a **minimum number of lead messages** before it evaluates at all, so a tag doesn't fire off a single "hey."

Tags can also match on either side of the conversation — what the *lead* said or what the *setter* said. A "Booking Link Sent" tag, for example, can fire whether the setter says "here's the link" or the lead asks for one.

### AI Tags vs. Pattern Match

Every tag runs in one of two modes:

| Use AI Tags when...                                           | Use Pattern Match when...                                          |
| ------------------------------------------------------------- | ------------------------------------------------------------------ |
| A lead could phrase the same thing 10 different ways          | A setter sends a specific, copy-pasted script or phrase            |
| Intent matters more than exact wording                        | You want to track one literal keyword or trigger word              |
| You're catching qualifying signals, objections, or engagement | You're tracking outbound opener scripts or a specific campaign CTA |

Pattern Match also supports matching a full sentence or phrase, not just a single keyword — useful for detecting that a lead's opening message matches a specific bio-link CTA.

### Create a tag with an AI prompt

**Web and iOS** (same nav path):

1. Go to `Settings > Tags`.
2. Click **New Tag** and give it a name and color.
3. Choose **AI Tag** as the type.
4. Write your prompt (see "Writing a strong AI tag prompt" below).
5. Set a minimum lead-message threshold if you want one.
6. Click **Save**.
7. Click **Retag leads** to sweep it across existing conversations — a new tag only evaluates new activity going forward until you do (see "Retag" below).

### Writing a strong AI tag prompt

Write the prompt like you're briefing a new hire whose only job is to apply this one tag. Be specific. State what qualifies and what explicitly does **not**.

**Weak:** "Tag if the lead seems interested."

**Strong:** "Tag as Hot Lead if the lead explicitly expresses urgency or readiness to buy after being nurtured. Examples: 'I'm ready,' 'where do I sign up,' 'how do I get access,' 'I want in.' Do NOT tag if they're just asking general questions."

Vague prompts are the single most common cause of "my AI tags aren't firing correctly." If a tag is under- or over-firing, the fix is almost always to tighten the prompt with more explicit examples and exclusions — not a platform bug. You can paste a draft prompt into Claude or ChatGPT and ask it to rewrite it as a specific, unambiguous tagging instruction.

For objection tags specifically, scope them to fire **only after a CTA has been sent**. A lead asking "how much is it?" before any offer has been made isn't an objection yet.

### Test before you go live

Use the tag tester to check a prompt against real conversations before turning a tag loose on your whole inbox. This catches obvious over- or under-firing before it affects your team's daily filters.

### Retag: sweep your existing leads

A tag only evaluates conversations going forward from the moment it's created or edited — it doesn't automatically reach back into your existing leads. Click **Retag leads** any time you create a new tag or meaningfully edit an existing prompt, so historical conversations get swept too.

### A starter tag library

Don't try to cover everything on day one — start with 6–8 tags and expand from there.

**Stage**

| Tag               | Type          |
| ----------------- | ------------- |
| Hot Lead          | AI            |
| Qualified         | AI            |
| Booking Link Sent | Pattern Match |
| Call Booked       | Pattern Match |
| No Show           | Pattern Match |

**Objection** (scope each to *only after a CTA has been sent*)

| Tag                   | Type |
| --------------------- | ---- |
| Objection — Price     | AI   |
| Objection — Time      | AI   |
| Objection — Trust     | AI   |
| Objection — Authority | AI   |

**Source**

| Tag                      | Type          |
| ------------------------ | ------------- |
| Source: \[Campaign name] | Pattern Match |
| Source: Outbound         | Pattern Match |
| Source: Story Reply      | Pattern Match |
| Source: Referral         | Pattern Match |

**Engagement**

| Tag                             | Type   |
| ------------------------------- | ------ |
| Budget Confirmed                | AI     |
| Needs Follow-Up                 | Manual |
| Language routing (e.g. "Dutch") | AI     |

### Tips

* Create one distinct tag per campaign, story sequence, or reel keyword — it's the easiest way to see which content is actually converting in Performance, filtered by tag volume.
* Add a few "lead archetype" tags mapped to your distinct customer profiles or objection patterns, so a setter opening a fresh conversation immediately knows how to approach that lead type.
* Review tag performance monthly. If a tag is firing too broadly, tighten the prompt rather than deleting it.
* A generic, manually-applied "Needs Follow-Up" tag is a simple, high-value addition alongside your AI tags.

### Common mistakes & fixes

| Mistake                                                      | Fix                                                                             |
| ------------------------------------------------------------ | ------------------------------------------------------------------------------- |
| Vague prompt ("tag if interested")                           | Rewrite with explicit qualifying phrases and explicit exclusions.               |
| Forgetting to Retag after creating or editing a tag          | Click **Retag leads** every time — otherwise only new activity is covered.      |
| Objection tags firing on pre-CTA questions                   | Scope the prompt explicitly to "only after a CTA has been sent."                |
| Two tags catching the same lead unintentionally              | Tighten both prompts' boundaries so they're mutually exclusive.                 |
| Using an AI tag for something that's really one fixed phrase | Switch it to Pattern Match — simpler and more predictable for literal triggers. |

### Limits & good to know

* Only Creator and Manager can create, edit, or Retag tags.
* A tag needs enough conversation history (and, if set, enough lead messages) to evaluate — it won't fire on an empty thread.
* AI tags need some baseline conversation data across your org to perform well, same as the other AI features in this module.

**Related:** Set Up Your AI: Setter Playbook & Creator DNA, Lead Summary, Spam Filter. Also see **Tags & Inbox Categories** and **The Leads Page** (Modules 02–03).