Website Conversion Audit

Give it page URLs. For each one it fetches that single page and reports what the served HTML shows about the path from "visitor on a phone" to "contact or booking": each finding quotes the markup, URL or HTTP status it was read from, so you can check it yourself in seconds. It never guesses: if a page cannot be audited, you get a refusal with the reason, free.

The report wording is written for appointment- and enquiry-driven local businesses (clinics, salons, trades, law firms).

What it checks (the complete list)

Every finding comes from one of these checks of the HTML the server sent:

• Mobile viewport tag: missing <meta name="viewport">, or one that disables pinch-zoom.
• Insecure form: a form on an https page whose action is http://.
• Tappable phone number: phone numbers in the page text but no tel: link anywhere.
• Wrong number behind a tel: link: link text shows one number, the href dials another (skipped when a call-tracking vendor script is present).
• Booking link: the primary booking link is followed once; a 4xx/5xx answer is reported as broken, a non-response as unreachable.
• After-hours path: no booking link, contact form, sms: link or known chat widget script in the HTML.
• Contact form shape (contact forms only, not newsletter/search): 6+ required fields, phone box without a number keypad, type="number" phone box that refuses ( ) + space, CAPTCHA.
• First tappable link is a mailto:.
• Stated reply time of a day or more, quoted from the page text ("we will respond within 48 hours").
• Opening hours not published in JSON-LD or recognisable page text.

Each finding is marked confirmed or partial. Absence findings are partial (and never "blocking") when a tag manager or similar injector is present, because it can add things this method cannot see. The report lists that blind spot by name.

What it does NOT see

• Anything added by JavaScript in the browser: tag-manager widgets, client-rendered forms (Wix, Squarespace, Webflow forms are not judged as absent).
• Other pages of the site. One page per URL; no crawl, no sitemap.
• How the page looks or behaves on a device: no rendering, no screenshots, no speed or response-time measurement.
• Search rankings or SEO. No promises about either.
• How fast staff reply, what a caller hears after hours, Google Business Profile or reviews.
• Pages that block robots, need a login, or are not HTML.

Politeness rules (not options)

• One GET per URL, plus at most one follow of the primary booking link. No crawling, no retries, no form submits, no headless clicking.
• robots.txt is read before fetching (one request per origin per run, cached). If it disallows the path for * or for EveSiteAudit, the result is robots_disallow and you are not charged. Redirect hops and the booking link are checked too.
• Honest User-Agent: EveSiteAudit/1.0 (+<this actor's Store URL>; one-time audit fetch). Never a browser string.
• Private, loopback and link-local addresses (v4 and v6, also after redirects) are refused.
• Page HTML and contact details are never logged or stored; the dataset holds the report only.

Input

Field	Description
urls (required)	Page URLs, max 25 per run. Duplicates are removed.
maxUrls	Audit at most this many (default 25). Extra URLs come back as free over_max_urls refusals.

Output

One dataset item per URL: { url, finalUrl, ok, reason?, findings[], markdown }. Refusal reasons include robots_disallow, not_auditable:* (parked, redirect stub, directory page, too small), http_<status>, not_html, private_address:*, dns_failed:*, fetch_failed:*, timeout_*.

Price

Pay per event: audit-completed, once per URL that produced a report (ok: true). Refusals and failures are free. The price is set in the Apify Console by the publisher (see PRICING.md for the recommendation). If your spending limit is reached the run stops and the unpaid report is not delivered.

Sample output

Produced locally from a made-up fixture (tests/fixtures/clinic.html, host clinic.example), not a real business:

{
  "url": "https://clinic.example/",
  "finalUrl": "https://clinic.example/",
  "ok": true,
  "findings": [
    { "id": "no-viewport", "severity": "blocking", "confidence": "confirmed",
      "title": "The page has no mobile viewport tag",
      "evidence": "No <meta name=\"viewport\"> in <head> of https://clinic.example/" },
    { "id": "insecure-form", "severity": "blocking", "confidence": "confirmed",
      "title": "Your contact form submits over an insecure connection",
      "evidence": "<form action=\"http://clinic.example/submit\" method=\"post\">" },
    { "id": "dead-booking-link", "severity": "blocking", "confidence": "confirmed",
      "title": "Your \"Book Now\" link is broken",
      "evidence": "<a href=\"/book-appointment\">\n  → https://clinic.example/book-appointment\n  → HTTP 404" }
  ],
  "markdown": "# Lead Conversion Audit — clinic.example\n\n**Page audited:** https://clinic.example/ ..."
}

(Findings abridged: the fixture yields 9 in total. A parked page returns {"ok": false, "reason": "not_auditable:page_too_small"}.)

Run locally

ACTOR_CONTACT_URL (required: the Store URL used in the User-Agent) and optional FIX_LIST_URL (appends Want the full report re-checked after your fixes? <FIX_LIST_URL> to each report) are environment variables. npm test runs the tests; NODE_OPTIONS="--import ./dev/fixtures-preload.mjs" ACTOR_CONTACT_URL=https://apify.example/x apify run runs the actor against the local fixtures with no network.