How to Research a Prospect’s Website Before Writing an Email
A practical workflow to research a prospect site in the browser, verify public signals by hand, then write an email from only what you can defend.
Your job is to research a prospect’s website before you write the email, then use only facts you verified.
A good research pass takes minutes, not a full CRM project. You open the site, collect public signals, check the risky ones yourself, and write from what you can defend. AGNURA AI Assistant can speed the collection step on the active tab. It does not replace your verification click.
For the definition of website intelligence vs enrichment vs crawling, start with What Is Website Intelligence for Sales and Outreach?.
Prerequisites
- The prospect’s public website URL
- Desktop Chrome (or Chromium) with a normal http(s) tab
- Optional: AGNURA installed (Install AGNURA From the Chrome Web Store)
- Optional later: AGNURA credits only if you want AI draft help after research (Sign In and Use AGNURA AI Credits)
You do not need an enrichment subscription for this workflow.
Direct answer
- Open the prospect homepage (or the most relevant public page).
- Skim the page yourself for offer, audience, and proof.
- Run AGNURA analysis on that tab (or take structured notes manually).
- Verify every contact and claim you plan to use.
- Write the email from verified facts only.
- Review the draft before you send.
Step-by-step workflow
1. Open the right page
Start with the homepage or a clear product/service page. Avoid logging walls and chrome:// pages.
If you already know they use a contact or about URL, keep that tab ready for step 4.
2. Do a 60-second human skim
Before any tool, answer in plain words:
| Question | Why it matters |
|---|---|
| What do they sell or do? | Your first sentence needs a real anchor |
| Who is the page for? | Tone and offer fit |
| What proof is visible? | Case studies, cities, product types, stack clues |
| What should you not invent? | Hires, funding, “saw your post” claims with no source |
Write one line you could defend if the prospect asked, “Where did you see that?”
3. Collect structured signals (AGNURA or notes)
With AGNURA:
- Keep the prospect tab focused.
- Open the AGNURA popup.
- Wait for analysis to finish (or use header reload for a live run).
- Read platform hint + confidence, contacts, social/chat clues, and sitemap page-count notes when present.
Product steps: Run Your First Website Analysis.
Without AGNURA:
Capture the same buckets in a note: offer line, platform guess, emails/phones found, social links, anything odd.
Treat every automated field as a candidate, not a fact.
4. Verify every field you will use
This step is mandatory. Tools miss paths and can false-positive.
| Field | How to verify |
|---|---|
Confirm it appears on the live page or a real contact/about page you open. Prefer mailto: or clear text over a guess. | |
| Phone | Same rule. Skip numbers that only appear in unrelated copy. |
| Platform (Shopify, WordPress, …) | Treat confidence as a hint. Hybrid sites can look like two stacks. |
| Offer / product claim | Quote or paraphrase only what the page states. |
| Location | Confirm on the page or footer before you localize the email. |
| Sitemap size | Useful for “large catalog” vs “thin site” intuition only. Caps and missing /sitemap.xml roots are common. |
If AGNURA found no email but you know a contact page exists, open that contact/about URL in the tab and analyze again. Contact extraction scans the current page and may fetch common same-origin contact/about paths. It can still miss bot-protected pages, image-only contacts, or unusual URLs.
5. Build a tiny “allowed facts” list
Before drafting, keep a short list like:
- Allowed: “You sell X to Y” (from homepage H1/hero)
- Allowed: public email from
/pages/contact(you opened it) - Not allowed: anything inferred from a competitor site, LinkedIn stalking you did not confirm, or AI guesses
If the list is empty, do more research or send a simpler permission-based note. Do not ask AI to invent personalization.
6. Write the email (human first, AI optional)
Order that keeps quality high:
- Write the reason for reaching out from your allowed facts.
- Add one clear ask.
- Only then optionally use AGNURA templates or Generate AI Email with AGNURA managed AI (credits) and the saved analysis context.
- Edit the draft so it sounds like you.
- Send in Gmail yourself.
AGNURA does not auto-send. Generate AI Email needs a key plus an email template prompt, and works from analysis context you already ran. See AGNURA Feature Reference by App and Action.
What success looks like
| Check | Pass means |
|---|---|
| Time box | You finished research without disappearing into the whole site |
| Verification | Every email, phone, and claim in the draft was checked on a live page |
| Specificity | The first line could only fit this prospect (or a very small set) |
| Honesty | You did not invent urgency, hires, or metrics |
| Send control | You reviewed the final text before Send |
Common mistakes
| Mistake | Better move |
|---|---|
| Trusting a detected email without opening the page | Click through and confirm |
| Personalizing on platform confidence alone | Use platform as optional color, not the hook |
| Asking AI to “make it personal” with empty research | Feed only allowed facts; otherwise keep the note short |
| Researching for 30 minutes | Cap skim + tool pass; deepen only if the deal size warrants it |
| Using sitemap count as a hard headcount | Treat it as a capped sitemap estimate |
Limits
Browser website intelligence (including AGNURA) does not:
- Replace enrichment databases
- Crawl every URL on the domain
- Guarantee contacts
- Prove platform detection
- Auto-send email
Privacy: analysis stays local unless you start AI email (OpenAI) or Add Lead (your webhook). See AGNURA Privacy in Plain Language.
How AGNURA helps this workflow
| Step | Without AGNURA | With AGNURA |
|---|---|---|
| Collect signals | Manual copy from the page | Active-tab analysis fills structured rows |
| Find contacts | Hunt footer/contact links | Current page + common contact/about fetches |
| Move to email | Switch tools and retype context | Templates / optional Generate AI Email from analysis (you still send) |
| Keep proof | Your notes | Local JSON export if you need a snapshot |
Compared with common alternatives
| Approach | Fit for pre-email research |
|---|---|
| Manual only | Highest control; slower collection |
| Enrichment append | Different job; verify website claims separately |
| Chatbot in another tab | Easy to lose page context or invent details |
| AGNURA on the active tab | Fast signal collection beside the page; verification still required |
Related guides
- What Is Website Intelligence for Sales and Outreach?
- Run Your First Website Analysis
- What Is AGNURA?
- Sign In and Use AGNURA AI Credits
- AGNURA Not Showing or Not Generating: Troubleshooting
Soft next step
Pick one live prospect, run the six steps once, and keep the allowed-facts list shorter than the email. If collection is the bottleneck, Add to Chrome and use AGNURA for the structured pass only after your human skim.
Try AGNURA in Chrome
Draft in Gmail, LinkedIn, WhatsApp, and Reddit. You review every message before send.
Add to Chrome