Kovik SEO Audit Engine

What it does

The engine fetches pages from your site the way a search engine does — one request at a time, same host only, respecting robots.txt — parses the HTML, and measures what is actually on the page. Nothing is assumed and nothing is estimated from a third-party index. Every line in the report comes from a measurement taken during that crawl.

It then scores five categories, ranks the failures by severity, and writes a PDF you can send to a client as it is.

Install

Python 3.9 or newer. Four libraries, no API keys.

pip install requests beautifulsoup4 lxml reportlab matplotlib

Download kovik_audit.py and requirements.txt from the repository, then install:

pip install -r requirements.txt

Run an audit

The shortest version — crawl up to ten pages and write a PDF next to the script:

python kovik_audit.py https://client-site.in --out client-audit.pdf

The terminal prints the crawl, the score for each category and the first six findings while the PDF is being written:

KOVIK · SEO AUDIT ENGINE

  [1/10] https://client-site.in/
  [2/10] https://client-site.in/services
  ...

  Overall score  84/100  (Good, with gaps)
  Technical        74
  On-page          81
  Content         100
  Structured data 100
  Performance      68

  3 failing · 3 to improve
    FIX  [critical] HTTPS on every page
    FIX  [high    ] XML sitemap present

  PDF  → client-audit.pdf
  Done in 1.6s

All options

What it checks

Twenty-six checks across five weighted categories.

A pass scores 1, an improvement scores 0.5 and a failure scores 0. Each check carries its own weight inside the category, so a missing title hurts more than a missing lang attribute. Categories are then weighted into the overall score out of 100.

Reading the report

The PDF is six pages, in this order:

1. Cover — site, date, score ring, what is inside.
2. Executive summary — score card, category chart, the five things to fix first.
3. Findings — grouped critical to low. Each one states what was measured, which pages were affected, and what to do about it.
4. Page-by-page table — status, page score, title and description length, H1 count, words, missing alt text, response time, page weight.
5. Method and limits — what the audit does not cover.

Every page carries a diagonal watermark and a footer line, so a report that leaves your inbox still says where it came from.

JSON and CSV output

Useful when you want to store the result, chart it over months, or feed it into a client portal.

python kovik_audit.py https://client-site.in --pages 10 --out audit.pdf --json audit.json --csv audit.csv

The JSON carries the crawl summary, the scores, every check with its status and fix text, and a row per page. The CSV carries the issue list only — severity, category, check, status, what was found, what to do, which pages.

Your own branding

python kovik_audit.py https://client-site.in --pages 12 --client "Sharma Dental Clinic, Hajipur" --out sharma-dental.pdf

Colours, the watermark text and the footer line live at the top of the report module — four constants. If you white-label the engine for another agency, that is the only file you need to touch.

Limits

• One crawl, from one location, at one moment. Response times are a signal, not a benchmark — run it twice before quoting a number.
• No backlink data, no search volume, no rank tracking, no Core Web Vitals field data. Those need Search Console, Analytics and a rank tracker.
• JavaScript-rendered content is not executed. A site that builds its content in the browser will look almost empty to the crawler — which is worth reporting on its own.
• It crawls one host. Subdomains and external pages are counted as outbound links, not crawled.

Troubleshooting

“Could not fetch the site”

Check the URL includes https://, that the site is reachable from your network, and that a proxy is not intercepting the request. On a corporate network, set HTTPS_PROXY before running.

Every page comes back with zero words

The site is rendering its content with JavaScript. Try a server-rendered page, or treat the empty result as the finding it is.

PDF looks wrong, fonts missing

The report uses Helvetica, which ships with reportlab. If you have edited the styles and introduced a custom font, register it with pdfmetrics.registerFont first.

Questions