Skip to content
Rezultra®
Request a callback

Home / Tools / Methodology

How we measure

We publish our method so you can judge our numbers, and hold us to them.

The SEO audit

We download your homepage, then up to nine more pages found in your navigation and sitemap, the way a crawler does, with our own clearly labelled user-agent (RezultraBot). We respect robots.txt and fetch three pages at a time. We don’t run JavaScript, because most AI crawlers don’t either.

Each check has a status, an impact (high, medium, low) and an effort estimate. Category scores weight checks by impact; AI readiness and indexing count 1.5× in the overall score because they decide whether you can be found at all. Site-wide checks (such as the robots.txt rules for search and AI crawlers, the sitemap, HSTS and the favicon) are read from your homepage and appear once in the issue list, not once per page. They still count in every page’s score, because they affect every page. Notes and llms.txt are never scored.

Your scores, the issue list with page counts and the page list are free. How to fix each issue, the pages each one affects, the page-by-page detail, the PDF and the share link unlock when you send the short form; the SEO audit is the only tool that holds anything back.

AI crawler tests

We evaluate robots.txt for 19 crawlers using RFC 9309 and Google’s rules: a crawler uses the group naming it (falling back to *), the longest matching rule wins, and Allow wins ties. Then we send one request as each of 12 of them (OAI-SearchBot, GPTBot, ChatGPT-User, Claude-SearchBot, ClaudeBot, PerplexityBot, Perplexity-User, Googlebot, Bingbot, Applebot, CCBot, Meta-ExternalAgent) with its published user-agent from Cloudflare’s network and report what your server returned. That is a simulation: real crawlers come from their own IP ranges, and some firewalls verify them by IP. The other 7 (Claude-User, Google-Extended, Applebot-Extended, DuckAssistBot, Meta-ExternalFetcher, Amazonbot, Bytespider) are checked against robots.txt only; Google-Extended and Applebot-Extended are robots.txt tokens with no crawler of their own.

  • AI search & answers: OAI-SearchBot, Claude-SearchBot, PerplexityBot, Googlebot, Bingbot, Applebot, DuckAssistBot, Amazonbot
  • Fetches on user request: ChatGPT-User, Claude-User, Perplexity-User, Meta-ExternalFetcher
  • Model training: GPTBot, ClaudeBot, Google-Extended, Applebot-Extended, CCBot, Meta-ExternalAgent, Bytespider

The robots.txt rules we suggest repeat your private paths in every group that names a bot, because a bot with its own group ignores the * group.

Social benchmark

YouTube. Channel and video data come from YouTube API Services (the YouTube Data API, with Rezultra’s own API key; you don’t sign in). We show a channel’s subscribers, total views and video count, and the publish date, views, likes and comments of its latest 15 uploads, exactly as YouTube returns them, with the time we fetched them. YouTube rounds subscriber counts and channels can hide them; when YouTube gives no number we show “—”, never zero. Since 24 August 2026 YouTube counts a view as soon as a video starts playing, including autoplay. YouTube’s developer rules don’t allow new metrics calculated from YouTube data without Google’s approval, so we don’t show engagement rates, scores or rankings for YouTube. Use of this tool is subject to theYouTube Terms of Service; Google’s Privacy Policy is at google.com/policies/privacy.

Website checks. From your homepage we check whether your site links to a YouTube channel, whether that channel is listed in your structured data (schema.org sameAs), whether the two point to the same channel, whether the link uses a durable address (an @handle or channel ID rather than an old /c/ or /user/ link) and whether the channel exists on YouTube. These checks are Rezultra’s own analysis, not YouTube data. We never fetch YouTube pages; channel data only comes through YouTube API Services.

Instagram works from numbers you type in: your followers, the average likes and comments per post over your last 10–12 posts (leaving out paid or boosted posts), and how many posts you made in the last 30 days. Engagement rate is average likes plus comments per post divided by followers; posting pace is posts in the last 30 days × 7 ÷ 30. We compare them with 0.30–0.48% per post, per follower and a median of 3.7 posts a week for brands (sources: Quid (Rival IQ) 2026 median, Socialinsider 2026), and combine the two into a 0–100 score (engagement counts for 65%, posting for 35%). If your numbers give an engagement rate above 25% we ask you to check them, since likes and comments should be averages for one post. The calculation runs in your browser: there’s no Instagram login and nothing is fetched from Instagram. Accounts you add for comparison are worked out the same way from the numbers you enter. We don’t ask for post dates, so we never call an account inactive; the posting pace covers it.

Facebook is coming soon.

This shows your own channel’s public numbers and scores the Instagram numbers you enter; it isn’t social listening. Listening tracks what other people say about your brand across the web, which needs paid data feeds; ask us if you need it.

Speed

Real-user data comes from Google’s Chrome UX Report (75th percentile, last 28 days, with 25 weeks of history). Lab data comes from a Lighthouse run via PageSpeed Insights. Thresholds: LCP ≤ 2.5 s, INP ≤ 200 ms, CLS ≤ 0.1.

Your data

  • The tools run on Cloudflare, like the rest of rezultra.com: the pages, the checks and the lead form all go through Cloudflare’s network, and the quick check before your first run is Cloudflare Turnstile. We fetch the website you enter to run the checks.
  • To keep repeat checks fast, results are kept for a short time in Cloudflare’s cache, in the data centre that served you: Core Web Vitals real-user data for 6 hours and lab tests for 1 hour; YouTube channel numbers for 6 hours; AI crawler tests and the page details behind the schema validator, SERP preview and llms.txt generator for 10 minutes. The SEO audit itself isn’t cached, but the parts that unlock with the form are kept for 2 hours, so they can be shown when you send it. A robots.txt file loaded into the tester may be kept by your browser for 5 minutes.
  • We don’t add YouTube data to the lead sheet: for the social benchmark it only gets the website and channel links you entered and, if you ask us to get in touch, the Instagram numbers you typed in for your own account (labelled as entered by you). Instagram numbers you type in are otherwise worked out in your browser and not sent to us.
  • When you send the form (to unlock the SEO audit report, or to ask us to get in touch from the SEO audit or the social benchmark), your name, email, phone, company, the service you chose and any note, with the tool, the site you checked and a short summary of the result, go from your browser to our server on Cloudflare. It passes them to Rezultra’s lead sheet in Google Sheets (through Google Apps Script) so our team can follow up, without keeping a copy; campaign details from your visit, such as UTM tags and ad click IDs, go with them. See our privacy policy.
  • Speed checks send the page address to Google.
  • Pasted code in the schema validator and robots.txt tester never leaves your browser.

Known limitations

  • The free audit covers 10 pages; a full crawl is available on request.
  • We don’t render JavaScript, so content added by scripts is reported as missing.
  • Bot simulations can be blocked by firewalls that would let the real crawler in.