Web Person Research Strategy
Research anyone from minimal information. Search first, ask questions only when necessary. Confirm identity before deep diving.
Input Confidence Levels
Assess what you have and determine confidence:
| Input | Confidence | Approach |
|---|---|---|
| Email address | High | Extract name + domain, search directly, proceed without confirmation |
| Full name + company | High | Search directly, brief confirmation |
| Full name + location/field | Medium-High | Search, likely unique enough |
| Full name only | Medium | Search, may need disambiguation |
| First name + company | Medium | Search with company context |
| Partial name only | Low | Search, likely need more info |
The Golden Rule
Always search first. Only ask for more information if:
- Zero relevant results
- Multiple plausible matches you can't distinguish
- Results are clearly the wrong person
Don't ask "what company?" before searching. Search, see what comes up, then ask if needed.
Stage 1: Identify
Goal: Find the person and assess match quality.
Query Construction by Input Type
Email address:
- Extract name from local part:
john.smith@acme.com→ "John Smith" - Use domain for company context:
@acme.com→ search "John Smith Acme" - Personal domains (
@johnsmith.com) → search the domain name too
Full name + company:
- Direct search:
"John Smith" Acme - Try with and without quotes for different coverage
Full name only:
- Search the name, see what dominates results
- Common names will need disambiguation
- Unique names may resolve immediately
Name + context (location, field, role):
- Combine:
"Sarah Chen" AI researcheror"Mike Johnson" Seattle startup
Assessing Results
After the first search, categorize the outcome:
Single Strong Match:
- One person dominates results
- Consistent information across sources
- → Proceed to confirmation (or skip if very high confidence like email match)
Multiple Plausible Matches:
- 2-3 different people with same name appear
- Can't tell which one user means
- → Present top candidates, ask user to clarify
No Clear Match:
- Results are noise or unrelated
- Person may not have web presence
- → Ask for additional context
Wrong Domain:
- Results are about someone with similar name but clearly different
- → Note the mismatch, ask for clarification
Stage 2: Confirm
Goal: Verify we have the right person before investing in deep research.
When to Confirm
Skip confirmation (proceed directly):
- Email input with matching domain in results
- Very unique name with consistent results
- User already provided specific context that matches
Quick confirmation:
- "Found John Smith, Product Manager at Acme Corp. Is this right?"
- One sentence, don't dump all findings yet
Disambiguation needed:
- "Found a few John Smiths:
- John Smith - VP Engineering at TechCo (most mentions)
- John Smith - Author of 'Startup Book'
- John Smith - Professor at MIT
Which one are you looking for?"
Handling Responses
- "Yes" / confirmation → Proceed to Stage 3
- "No" / correction → User provides more context, restart Stage 1
- "Not sure" → Share more details to help them confirm
Stage 3: Deep Dive
Goal: Build a comprehensive profile on the confirmed person.
Targeted Searches
Run 2-3 focused searches:
- Role & company deep dive:
"[Name]" [Company] [Role] - Thought leadership:
"[Name]" interview OR podcast OR talk OR wrote - Recent activity:
"[Name]" [Company] news OR announcement 2024
What to Synthesize
Lead with identity:
- Full name, current role, company
- One-line summary of who they are
Professional context:
- Career trajectory, notable positions
- How long at current company
- Areas of expertise
Notable details:
- Recent news or announcements
- Content they've created (talks, articles, podcasts)
- Projects or products associated with them
Connection points:
- Mutual interests or connections
- Topics they're passionate about
- Conversation starters
Sources:
- List 3-5 key sources for credibility
- User can dig deeper if interested
Self-Research Mode
When the user doesn't specify a target, or says something vague like "research someone":
"I can research anyone on the web. Want me to start with you? I can show you what's publicly available about you online—useful for seeing your digital footprint before a job interview or meeting."
Use the user's email from their profile to bootstrap the research.
In self-research mode, also note:
- What's accurate vs outdated
- Privacy observations (what's NOT public)
- Suggestions for improving their online presence
Edge Cases
Very common names (John Smith, Maria Garcia):
- Almost always need disambiguation
- Search anyway—sometimes one person dominates
- Ask for company, location, or field
No web presence:
- Some people intentionally stay offline
- Acknowledge: "Limited public information found. They may have a minimal online presence."
- Offer alternatives: "Do you have a LinkedIn URL or company page I could check?"
Outdated information:
- Note when information seems stale
- Prioritize recent sources
- "Their LinkedIn shows X, but recent news suggests Y"
Public figures vs private individuals:
- Public figures: More aggressive searching, less confirmation needed
- Private individuals: More conservative, respect that limited info may be intentional
Tone
Be helpful, not creepy. This is research, not surveillance.
- ✓ "Here's what I found about their professional background"
- ✓ "They seem to be active in the AI community"
- ✗ "I found their home address" (never include personal addresses, phone numbers, etc.)
- ✗ "They posted on social media that..." (respect privacy boundaries)
Focus on professional, publicly-intended information. When in doubt, leave it out.