To find company websites from company names in bulk, import your list into Datablist and run the Find Company Domains from Company Names enrichment. I recommend starting with the company dataset and using Google as a fallback. Review the confidence column, then send only ambiguous or unresolved names to the AI Agent with extra context.
The result is a table with one company domain and full website URL next to each matched company name. You also get a confidence level and a status for every processed row.
📌 Workflow summary
Input: company name, plus country when available
Output: company domain, full URL, confidence, and status
Recommended sequence: company dataset → Google fallback → AI Agent for ambiguous rows
Verified enrichment cost: 1 credit for a successful dataset lookup or 2.5 credits when Google is used. AI Agent cost varies with the research needed.
Quick links
- See the expected result
- Prepare the input table
- Run the bulk workflow
- Reuse the AI prompt
- Check the measured example
- Validate the websites
Expected result
Suppose you start with three company names and their countries:
| Company Name | Country |
|---|---|
| Drooms | Germany |
| EVACO GmbH | Germany |
| tanner ag | Switzerland |
After the enrichment, the same rows contain the fields you need for email finding, company enrichment, deduplication, or CRM import:
| Company Name | Company Domain | Company Full URL | Domain Confidence | Status |
|---|---|---|---|---|
| Drooms | drooms.com | http://www.drooms.com | High | success |
| EVACO GmbH | evaco.de | http://www.evaco.de | High | success |
| tanner ag | mtannerag.ch | https://mtannerag.ch/en/ | High | success |
These examples come from the run shown later in this guide. The domain is the clean identifier you normally pass to other enrichments. The full URL is useful when you need to visit or scrape the website.
Confidence is a review signal, not proof. A high-confidence match can still be wrong when two companies share a name, so I always inspect a sample before using the domains downstream.
Prepare your input table
The enrichment only requires a company name. A country helps separate businesses with similar names and improves local Google results.
I prefer this input shape:
| Column | Required? | Example | Why it helps |
|---|---|---|---|
| Company Name | Yes | Summit Logistics GmbH | Main lookup value |
| Country | Recommended | Germany | Narrows dataset and Google results |
| City | Optional for AI | Hamburg | Helps resolve common names |
| Industry | Optional for AI | Freight forwarding | Confirms the website content |
| Description | Optional for AI | Cold-chain logistics provider | Gives the AI Agent more context |
Clean obvious noise before the lookup. Remove notes, phone numbers, and legal text unrelated to the name. Keep legal suffixes such as GmbH or Ltd when they help distinguish the business. If your source is messy, follow the company name cleaning workflow first.
Weak inputs include generic labels such as Marketing Agency, branch names without the parent company, and names copied with addresses in the same cell. No lookup method can identify the right business from context it never received.
Choose the right lookup method
This task works best as a waterfall. Run the cheapest method first, then spend more only on rows that need it.
| Method | Best for | Cost | Main limitation |
|---|---|---|---|
| Company dataset | Known companies and clean names | 1 credit per successful lookup | May miss small or new companies |
| Dataset + Google fallback | Mixed lists where coverage matters | 1 credit for a dataset match, 2.5 credits when Google is used | Search may select the wrong company for ambiguous names |
| Google only | Live search for every row | 2.5 credits per lookup | Charges whether or not a website is returned |
| AI Agent | Ambiguous names with country, city, industry, or description | Usage-based | Slower and still requires review |
For most lists, I use Company dataset + fallback on Google. It handles easy rows cheaply and searches the web only when the dataset has no match.
The domain finder returns High, Medium, Low, or Very Low confidence. Multiple sources agreeing on a domain can produce high confidence. Google title matches and fuzzy name similarity also affect the score. Use the value to decide which rows you accept, review, or research again.
Find company websites in bulk with Datablist
Step 1: Import the company list
Create a Datablist account, start a collection, and import your CSV or Excel file.
Check the imported columns before running anything. The company name should be a text property. Keep the country in its own property if each row can have a different country.
Step 2: Open the domain finder
Click Enrich, open the Companies category, and select Find Company Domain from Company Name.
Step 3: Configure the lookup workflow
Choose one of these workflow settings:
- Use Companies Dataset + Fallback on Google: my default for a mixed list
- Use only Companies Dataset: the lowest-cost first pass
- Use only Google: live search on every row
Set Target Country when the whole list covers one country. Enable Define target country per item when your file already has a country column.
When Google is enabled, two settings can improve difficult searches:
- Accept non-root websites: allow results such as
platform.com/company-name. Leave this off unless company sites are hosted on a shared platform. - Skip following domains: exclude an industry directory that keeps appearing. Datablist already skips common directories and social networks when scoring results.
Datablist searches with the selected country and scores the top results using factors such as result position and title similarity. If you leave the country empty, Google uses the United States location by default.
Step 4: Map the input and outputs
Map Company Name to the property containing your names. If you enabled a country per row, map Target Country too.
Add these output properties:
- Company Domain
- Company Full URL
- Domain Confidence
- Status
I keep both domain and full URL. The domain is better for matching and deduplication, while the full URL preserves the exact page selected by the enrichment.
Step 5: Test a sample, then run the list
Run the enrichment on the first 10 rows. Check whether the websites belong to the intended companies, especially for short or common names. If the sample looks good, run it on the remaining view.
Use filters to control the waterfall. For example, filter for an empty website before a second pass, or filter for Low and Very Low confidence before AI research. This prevents you from paying to process rows that already have an acceptable result.
Use AI for ambiguous company names
Google can find a plausible website and still pick the wrong company. Names such as United, Pioneer, or Summit Solutions need context.
For those rows, open Enrich, choose AI Agent, and load the Advanced Company Name to Website template. Map the company name plus any country, city, industry, address, or description properties you have.
Reusable AI Agent prompt
Use this prompt as a starting point and replace each /ColumnName with the matching property from your collection:
Find the official website for the company below.
Search the web for possible matches, then visit the likely websites and compare their content with the supplied context.
Return the corporate website on its root domain. Exclude company directories, legal listings, map listings, shops, social networks, parked domains, and news articles. If no result matches the context, or if you are not confident, return an empty website.
Company name: /Company NameCountry: /Country
City: /City
Industry: /Industry
Description: /Description
Ask for three outputs:
- Company Website: the selected URL
- Result Explanation: the evidence linking the website to the row
- Confidence Score: a numeric review signal
The instruction to return an empty value matters. Without it, a research agent can feel pressured to choose the least-wrong result when no official site exists.
Run this method only on unresolved or uncertain rows. AI research costs vary because one company may require more searches and page visits than another. Set a research-iteration limit before the run to control spend.
Measured result from an example run
The example below gives a concrete result rather than a projected benchmark:
- Input rows processed: 18
- Websites returned in the displayed rows: 18
- Reported runtime: less than one minute
- Credits used: 21
- Displayed confidence: High for all 18 rows
This run shows 100% output coverage for a small sample. It does not prove 100% accuracy, and it should not be treated as a speed or cost guarantee for another list. Coverage depends on the companies, workflow setting, and Google fallback usage. Accuracy requires checking whether each URL belongs to the intended company.
For your own run, record four numbers:
- Coverage: rows with a non-empty website divided by processed rows
- High-confidence coverage: rows marked High divided by processed rows
- Sample accuracy: correct websites divided by manually checked websites
- Cost per accepted website: credits spent divided by websites that passed review
These measurements let you compare workflow changes without confusing a returned URL with a correct result.
Failure cases and limitations
No company name lookup is perfect. Watch for these cases:
- Several companies share the same name. Add country, city, industry, or a description and use the AI Agent.
- The company has no website. Google may still return a directory or a similarly named business. Accept an empty result instead of forcing a match.
- The website uses another brand name. The domain may look unrelated to the legal company name. Check the site footer, About page, address, or legal notice.
- A regional page ranks above the main site. Set the target country and decide whether you need the local or global domain.
- The company lives on a shared platform. Enable non-root websites only when you expect URLs with a path.
- Directories dominate the results. Add the recurring directory domains to the skip list and rerun the uncertain rows.
- The source name is dirty. Clean names mixed with locations, notes, or registration text before lookup.
- Confidence is not verification. The standard enrichment scores name and search signals. It does not prove ownership of the website.
Google results are cached for 48 hours. If you change Google scoring settings during that period, Datablist can reuse the saved results. This is useful for testing exclusions, but a rerun may not reflect a new live search.
Validation checklist
Before exporting or running another enrichment, check the following:
- The company domain opens a live website.
- The site name, products, or services match the company in the row.
- The country, city, address, or legal details match when available.
- The result is an official website, not a directory, marketplace, social profile, or news article.
- High-confidence rows passed a manual sample check.
- Medium, Low, and Very Low rows were reviewed or sent to the AI Agent.
- Empty results stayed empty when no reliable official website was found.
- Duplicate domains were checked for parent companies, subsidiaries, or wrong matches.
- Coverage, sample accuracy, runtime, and credits were recorded.
Once the website column passes review, you can use it for an email finder, company enrichment, or a clean CRM import. If your starting point is a LinkedIn company URL instead of a name, use the LinkedIn company page to website workflow.
Frequently asked questions
What is the minimum input needed to find a company website?
A company name is enough to run the domain finder. Add a country whenever possible. For ambiguous names, city, industry, address, or a short description improves the AI research step.
What does the domain finder return?
It can return the company domain, the full website URL, a Domain Confidence value, and a status. The status is success, not_found, or empty_name.
How much does a company name lookup cost?
The company dataset costs 1 credit per successful lookup. Google costs 2.5 credits per lookup, including searches with no returned website. AI Agent usage depends on the research performed, so reserve it for rows that need context.
Should I accept every high-confidence result?
No. High confidence means the available matching signals are strong. Review a sample and check ambiguous names before sending the websites into outreach, CRM, or further enrichment workflows.