How AGNURA Uses LinkedIn Profile Context, and What It May Miss
What LinkedIn profile fields AGNURA loads for Write Reply and invite notes, how messaging-only fetches work, what lazy/private views miss, and what gets sent to managed AI.
When AGNURA drafts a LinkedIn message or invite note, it often uses recipient profile context. That context is best-effort. It is not a complete CRM profile and not a guarantee that every About line or birthday made it into the model.
This article explains what AGNURA AI Assistant tries to load, how messaging-only chats fetch more detail, what commonly goes missing, and what leaves your browser when you click Generate.
Hub map: LinkedIn Writing Workflows: Messages, Invitations, Comments, and Posts. How-to for messaging: How to Write a LinkedIn Reply From Profile and Thread Context.
Direct answer
For Write Reply (messaging), AGNURA can send OpenAI:
- Your LinkedIn reply prompt
- Optional What I want to say
- Optional knowledge base
- Recipient profile text (when loaded)
- Conversation thread (when available)
For Write Invite Note, profile (+ recent Activity when visible) matters; there is no messaging thread. Rewrite actions do not reload profile by design.
If profile fields are empty, private, or not yet scrolled into view, the draft will not magically know them. Edit the profile field or write those details yourself.
Why this matters
People assume “AI saw the profile” means the same thing as “I opened the full profile and read it.” Those are different.
| Expectation | Reality in AGNURA |
|---|---|
| Full LinkedIn profile graph | Heuristic scrape of visible/fetchable fields |
| Always includes birthday and Activity | Only when Highlights/Activity are available to the scrape |
| Works identically from messaging-only UI | May need Voyager/iframe/background tab fetch; still best-effort |
| Rewrite also re-reads the profile | Rewrite is draft-only |
Trust the modal preview. If the profile accordion looks thin, the model is thin too.
What profile fields AGNURA tries to use
When loading succeeds, the recipient profile summary can include:
| Field | Notes |
|---|---|
| Name | From profile page and/or messaging header |
| Headline | When visible |
| Company / location | When visible |
| About | When parsed from the profile |
| Top skills | When visible |
| Experience | When visible in supported markup |
| Birthday | From Highlights when shown |
| Top Activity posts | Up to 3 recent Activity items when the profile feed carousel is on the page (invite notes also use this for personalization; authored activity preferred over reposts when filtered) |
Parsing targets LinkedIn SDUI-style markup (top card, About, Experience, skills). LinkedIn UI changes can break selectors until the extension is updated.
How profile context is loaded
From an open profile or rich messaging header
If you have the profile page open, or messaging already shows enough header detail, AGNURA reads what is in the DOM.
From messaging when only a profile URL is exposed
Then AGNURA may try, using your existing LinkedIn session:
- LinkedIn’s internal Voyager API paths
- A hidden iframe profile load
- If needed, a brief background profile tab scrape
All of that is still best-effort. It is not a LinkedIn partnership API for AGNURA, and it can fail silently into a thin profile field.
Birthday steering (Write Reply)
| Signal | Draft steering |
|---|---|
| Birthday today shown | Model is steered to open with a birthday wish first |
| Recent / belated birthday shown | Model is steered to open with a belated wish first, then continue |
If birthday is missing from the scrape, do not expect that opener. Verify before send.
What it may miss
| Situation | What often goes missing |
|---|---|
| Lazy-loaded sections | Experience, skills, Activity until you scroll/open the profile |
| Private or restricted profiles | About, experience, Activity, birthday |
| Messaging-only thin UI | Everything beyond name/headline until a fetch path succeeds |
| Activity carousel not on page | Top 3 Activity posts |
| Highlights hidden | Birthday |
| DOM / SDUI redesign | Partial or empty profile until AGNURA updates selectors |
| You used Rewrite Message | Entire profile and thread (by design) |
Empty profile does not always block Write Reply: you still need at least one of profile or conversation thread. Profile-only works for new chats. Thread-only works when history exists.
Data handling (what leaves the browser)
| Moment | What happens |
|---|---|
| Open Write Reply / Write Invite Note | Profile may load into the modal locally |
| Click Generate | Prompt + knowledge base (if any) + profile (+ thread for Write Reply) go to OpenAI via AGNURA managed AI (credits) |
| Click Rewrite | Draft + rewrite prompt only; no profile reload |
| Insert / Send | Manual; AGNURA does not auto-send |
AGNURA does not sync that profile bundle to an AGNURA cloud account in the current product. Follow LinkedIn’s terms and your company policy for sending profile text to managed AI.
Privacy: AGNURA Privacy in Plain Language. Local vs remote AI: Local Settings vs Local AI: The Privacy Difference.
Options and trade-offs
| Approach | Upside | Trade-off |
|---|---|---|
| Open the full profile, then Write Reply | Richest DOM scrape | Extra tab |
| Messaging-only Write Reply | Faster | Higher chance of thin profile |
| Edit the profile accordion before Generate | You fix gaps | Manual effort |
| Paste a short true hook into What I want to say | Steers the draft | You must keep it honest |
| Skip AI when profile is empty and stakes are high | Safest | Slower |
Decision framework
| If… | Then… | Because… |
|---|---|---|
| Profile accordion shows only a name | Open the profile page, scroll, retry; or Edit/paste | Thin in → thin out |
| You need Activity-based personalization | Open the profile so Activity can load | Carousel must be available to scrape |
| You only want wording polish | Use Rewrite Message | No profile send on rewrite |
| Birthday wish appears but Highlights look wrong | Delete or rewrite before send | Steering can misfire on bad/missing data |
| The person is private / limited | Write from thread + what you truly know | Do not invent |
Checklist before you trust a “personalized” draft
- Expand Recipient profile in the modal.
- Confirm the facts you care about are present.
- Confirm Activity lines (if used) match the live profile.
- Fill What I want to say with your real point.
- Generate, then cut anything not supported by the preview or the live page.
- Insert and send yourself.
Where AGNURA fits
| Need | With AGNURA |
|---|---|
| Auto-load profile beside messaging compose | Best-effort Write Reply profile field |
| Invite notes with profile + Activity | Write Invite Note (200-character cap) |
| Avoid re-sending profile on polish | Rewrite Message / Rewrite Invite Note |
| Stay accountable | Manual insert; you send |
Compared with common alternatives
| Approach | Fit |
|---|---|
| Manual read + type | Highest accuracy; slower |
| Copy profile into a chatbot tab | You control paste; more friction |
| Enrichment databases | Different job; may not match the live profile |
| AGNURA profile context | Same-tab convenience with explicit miss modes |
Risks and limits
- Profile context is heuristic, not certified identity data.
- Session-backed fetches use your LinkedIn login context; use only on accounts you are allowed to operate.
- AI may still invent details not in the profile field. Your review is mandatory.
- No acceptance-rate or reply-rate promises from richer profile context.
Related guides
- LinkedIn Writing Workflows: Messages, Invitations, Comments, and Posts
- How to Write a LinkedIn Reply From Profile and Thread Context
- How to Write a LinkedIn Connection Note Under 200 Characters
- Human-in-the-Loop Writing: Why Drafting Is Not Sending
- AGNURA Privacy in Plain Language
Soft next step
Open Write Reply on one chat, expand the profile accordion, and compare it to the live LinkedIn profile before you Generate. Install AGNURA if you need the messaging helpers:
Try AGNURA in Chrome
Draft in Gmail, LinkedIn, WhatsApp, and Reddit. You review every message before send.
Add to Chrome