Website: tech stack, health, emails and phone numbers
Reachable's Website connector reads a company website the way a visitor does, 10 pages at most. For each row of a list, it returns the CMS and technologies, the HTTPS certificate status, the PageSpeed mobile score, the email provider, and the published emails and phone numbers, each with its source and read date.
By Alexandre Rastello · Published on
In this series: Connectors
Sources: Reachable, 150 websites of RGE-certified companies read by the Website connector, 12 November 2026, ADEME, list of RGE-certified companies (data.ademe.fr), updated 12 November 2026, Google, PageSpeed Insights API, CNIL, B2B email prospecting rules (in French) · Updated on
What is the Website connector?
The Website connector is a column: for each row of a list that has a domain, it reads the website and returns dated fields. The row can come from any source (Google Maps, a CSV, the French company registry) and does not need a SIREN number.
| Data publisher | The company itself, on its website; Google for performance |
| Access | Reachable's reader (10 pages at most, robots.txt respected), DNS, TLS certificate, Google PageSpeed Insights API |
| Coverage | Any reachable public website, in France and elsewhere |
| Input | A domain (example.com); the row may have no SIREN |
| Freshness | Read at run time; how often the same domain is reread is shown on the connectors hub, and each field carries its read date |
| Price per run | Shown on pricing and on the connectors hub, from the current price grid |
What data does the Website connector return?
Two columns share one pass over the site: Website (technologies, hosting, email, content) and Performance (PageSpeed Insights). The site is read once even when both columns run.
| Family | Fields | Source |
|---|---|---|
| Technologies | Technology list, CMS, e-commerce, analytics tools, cookie banner | Page code, headers and scripts (Wappalyzer rules) |
| Hosting | Hosting provider, CDN, HTTPS certificate valid or not, expiry date | DNS, IP address, TLS certificate |
| Email provider (MX), DMARC record present or not | DNS | |
| Content | Pages found (contact, legal notice, shop, blog, careers, booking), language, social networks, LinkedIn page | Website pages |
| Contact details | Generic emails, personal emails, phone numbers, mobile numbers | Website pages, as published |
| Performance | Mobile and desktop score out of 100, LCP, CLS, TBT, Chrome field data when Google has it | Google PageSpeed Insights |
Emails and phone numbers are returned as the company publishes them, personal addresses included, with the page each one appears on. Third-party details (hosting company, web agency) are left out.
Real sample: six websites from the RGE sample below, read on 12 November 2026 (source: Website connector and PageSpeed Insights).
| Domain | CMS | HTTPS certificate | Mobile score | Email provider |
|---|---|---|---|---|
| pavageaupastre.fr | WordPress | valid, 50 days left | 70 | Microsoft 365 |
| kheops-guerin.fr | Duda | valid, 59 days left | 63 | Microsoft 365 |
| archisen.fr | Wix | valid, 37 days left | 56 | Infomaniak |
| nicomat.fr | WordPress | valid, 37 days left | 77 | OVHcloud |
| durandfils-25.com | WordPress | valid, 86 days left | 48 | OVHcloud |
| etsdallemagne.fr | none recognised | valid, 60 days left | 100 | not read |
What does a real sample of small-business websites show?
Out of 150 websites of French renovation tradespeople, almost one in five no longer responds, and most of those that do are slow on mobile. On 12 November 2026, we read 150 websites drawn at random (fixed seed) from ADEME's public list of RGE-certified companies (the French energy-renovation quality label), with the Website connector. 122 responded; 28 (19%) were unreachable. Of the 122:
| Measure | Sites | Share |
|---|---|---|
| Valid HTTPS certificate | 106 | 87% |
| Invalid certificate (expired, wrong name…) | 8 | 7% |
| HTTPS connection failed (no HTTPS or unreadable certificate) | 8 | 7% |
| CMS recognised | 81 | 66% |
| of which WordPress | 54 | 44% |
| of which Wix or Duda (8 each) | 16 | 13% |
| DMARC record published | 39 | 32% |
Mobile performance (PageSpeed Insights, 110 sites measured, 12 without a result): the median score is 72 out of 100. 20 sites (18%) score 90 or more, 13 (12%) score under 50. For 74 sites (67%), the largest element of the page takes more than 4 seconds to appear in the lab test (LCP), the threshold Google rates as poor. Google only has Chrome field data for 5 of these sites.
After those CMSs come Jimdo (3), Simplébo (2), E-monsite (2), then Sulu, Weebly, Drupal and Plone (1 each). Only 4 sites run an online shop.
The date a website was last updated is not measured: our reader records no reliable date, and we do not publish an estimate. This point remains open.
Method: 150 connector reads, one per domain, on 12 November 2026 between 15:40 and 15:53 UTC, on the same sample as the email extractor study. An unreachable site has no page read successfully. The CMS is the one the Website column keeps. Shares are rounded to the nearest unit.
What prospecting signals can a website give you?
A website shows what a company has already bought and what it lacks: that is the basis of technographic prospecting.
- Unreachable site or no valid HTTPS: a web agency or hosting company has a concrete, dated reason to call.
- Certificate about to expire: 7 of the 106 valid certificates in the sample expired within 30 days.
- Given CMS: a WordPress plugin vendor or a Wix or Shopify partner targets sites on its platform; a list rerun on a schedule adds those that switch.
- Low mobile score: a quantified argument for a redesign or optimisation.
- No DMARC: an argument for an email or deliverability provider.
- Published contact details: the email or phone number to use, with the page it appears on.
These signals live in a list's filter: at each rerun, newly qualifying companies arrive by export or by webhook at the end of the run.
Which other connectors work well with the Website connector?
The Website connector needs a domain; other connectors supply the companies and the context. Three recipes:
- Google Maps + Website: the businesses of a trade in an area, then, for those that declare a website, its CMS, certificate and published contact details.
- RGE + Website: RGE-certified tradespeople in a French département, keeping those whose site is slow on mobile or lacks valid HTTPS, for a redesign offer.
- SIRENE + Website: companies of a sector and size from the French registry, matched to their website, to target those using a given technology.
For a single site, without an account, the free email extractor uses the same reader. Every connector and its price per run is on the connectors hub.
How do you use it from an AI agent (MCP)?
Your agent (Claude, ChatGPT, Cursor) adds the Website column to a list in plain language through Reachable's MCP server. Example request:
On my list of plumbers in the Alpes-Maritimes, add the Website column, keep those on WordPress with a mobile score under 50, and give me their published emails.
The agent returns a table: domain, CMS, certificate, mobile score, emails and phone numbers, each field with its source and read date.
What are the Website connector's limits?
The connector reads what the website shows a visitor; it guesses nothing and does not check it elsewhere.
- 10 pages at most, robots.txt respected: contact details published deeper, in an image or behind a click, are missed.
- Signature-based detection: a hidden CMS or a hand-built site is not recognised (34% of the sample).
- No last-updated date for the site, nor content age.
- PageSpeed Insights may return nothing for a slow site or one that blocks Google (12 of 122 in the sample); lab scores vary between runs.
- Shared reads: a domain read recently is not reread at every run; the frequency is on the connectors hub and each value shows its read date.
- Network failure: a field that DNS or the certificate could not be read for is marked unread, neither written nor billed.
- Matching: a SIREN read in the legal notice is not taken as certain. Matching the site to a company carries a method and a confidence score; below the list's threshold (0.95 by default), columns that need a SIREN do not run. See our methodology.
- Personal data: a published personal email is still personal data; its use follows the GDPR.