How onboarding reads a link or a social handle, shows what it found, and lets the person keep or turn down each line
During onboarding, a link or a social handle can be read before the conversation moves on. The person sees a card in the chat: first "Doing the homework...", then "Here's what I found about you." Every line starts as theirs. They turn down what is wrong with Not me and finish with Done. Only the lines they kept are saved, and the agent continues from there.
While the card is open there is nothing to type. The card is the only way forward. For the rest of setup, see Agent Setup Onboarding.
The visitor starts as a guest. Signing in comes near the end.
A link is never required. A short description of who they are is enough to move on, and that description is not researched.
The same card is used on talent, brand, and agency onboarding. Talent starts with a handle. A brand or agency starts with its website. A handle mentioned after that website step, such as an example of talent, is not researched as the company. See Talent, Brand, and Agency.
A link can arrive at any point, not only on the first question. Each new link gets its own card.
They send a link or a handle. The reply shows Doing the homework... Typing is disabled until they press Done.
The same card fills in with one line per finding:
| Title | What it shows |
|---|---|
| Who you are | Name, and a role when one was found |
| About | A one-line description |
| Audience | Who they reach |
| Region | Where they are strongest |
| Recent work | Up to three pieces of recent work |
| Found on | Where they were found, with a link. When the research also found the handle or a follower count, the line carries them: "Instagram — @lebron, about 4.2M followers" |
Each line has Not me (it becomes Undo). The card has one Done. There is no Skip. If nothing is them, they press Not me on every line.
They press Done. Turned-down lines stay struck through. Typing comes back, and the agent asks the next question without them having to send another message.
The card stays in the chat as they left it, including after a reload. A later link adds a second card. The first one does not change.
If the page could not be opened, the card says "I couldn't access that link." and the conversation carries on. The agent does not repeat that.
The homework starts when the message has something to open: a web address, or a handle such as @janekim. "I'm a designer in Tel Aviv" does not start it.
Sending the link only remembers it. The card appears when the agent takes that link up in its reply. If the agent never does, there is no card and no homework.
LinkedIn needs the full profile link, like linkedin.com/in/someone. @someone is not a LinkedIn address. On Instagram, TikTok, Facebook, Threads, and X, the handle is enough.
A normal website is read as a web page. A social profile or a handle is read as a profile, through Bright Data. That needs BRIGHTDATA_TOKEN on the server. Without it, a LinkedIn link says it could not be opened. The public LinkedIn page is not used instead, because it mixes the person's card with other people's posts.
The name they typed stays, unless the profile name starts with it and is longer. "Elad" against "Elad Itzhakian" shows the full name, even with no job title. A different name does not replace what they typed.
LinkedIn often has no headline and no job history. The card uses the name, company, and place that did come back. It does not invent a job title.
If a page cannot be read, the card does not make up a biography. Homework gives up after about two minutes. If it is still going after that, the next look treats it as failed and typing comes back.
Done saves the lines they left on:
A guest has no account yet, so this waits with the conversation. Sign-in copies it onto the account.
Turning down "Found on" keeps that page off the profile. It does not delete the link from the account's saved links.
A name sent in the same message as a link ("neymar, here is my link …") is saved as just the name.
| What happened | What they see |
|---|---|
| The page cannot be opened | "I couldn't access that link." The chat moves on. |
| Homework fails or hangs | The same message. Typing comes back. |
| Every line turned down | The card stays, all struck through. Nothing from it is saved. |
| They reload while it is still reading | The card picks up where it left off. |
| They reload after Done | The card is still there, read-only. |
| They send the same link again | It is researched again. |
| They want to correct it after Done | They say so in the chat. The card is already closed. |