Argos: audit a web page and prioritise fixes
Start from a public URL, read the evidence and coverage in an Argos report, then prepare a correction you can verify.
Choose the page and question to examine
You are taking over a website and need to decide which correction to address first. An important page seems slow, but you do not yet know what other technical defects it has. I created Argos to bring these initial signals together in a report you can examine with your team.
The tool analyses a public URL for performance, security, SEO and accessibility. From its home page, you can start a free analysis without an account. I would begin with a page tied to a specific use, such as a product page visitors need to understand before contacting the business. The result concerns that URL; other pages and journeys behind a login must be examined separately.
After entering the HTTP or HTTPS address, you can follow the checks' progress. The time required depends on the page and available services. Bot protection can prevent data collection: you must account for that absence when reading the report.
Read coverage before comparing scores
A score summarises the checks measured. Coverage indicates the weighted proportion of expected checks that could be measured, excluding those that do not apply. It helps establish whether a score rests on a still-partial diagnosis. An unavailable measurement does not automatically receive zero; insufficient coverage makes the score provisional.
Before comparing two reports, I would examine the method used and measurements available in each. If collection was missing on the first run, a score difference may come from the measured scope. To judge a correction, you also need to find the relevant finding and use comparable analysis conditions.
Move from a finding to a precise change
The report connects findings to observed evidence and recommendations. To make a decision, I suggest opening a finding and locating its evidence on the page. Which part of the journey is affected? Can the defect be reproduced? Does the suggested correction suit this content?
Argos's public example uses a fictional shop to explain this reading process. In particular, it shows an informative image without an alt attribute. Its report data is illustrative and does not come from a full analysis. It can teach you to examine HTML evidence without inferring a measured result on a real site.
<!-- Avant : image informative sans alternative -->
<img src="/produit/carnet.jpg" width="640" height="480">
<!-- Après : texte à adapter à l’image et à son contexte -->
<img src="/produit/carnet.jpg" width="640" height="480"
alt="Carnet à couverture bleue ouvert sur une page quadrillée">In this separate example, the text must be reviewed alongside the image and surrounding content. A decorative image would need an empty alternative. An automated check can detect a missing attribute; its presence after a fix is not enough to establish that the text helps the reader.
Write a task the team can verify
Severity helps identify findings to examine. Work order also depends on the effect on users and the site's constraints. A defect that prevents an essential use may take priority over an improvement more visible in the score.
For the previous image, a task could request a useful alternative on the affected page and a check of whether the same component serves other products. The report does not prove those other pages have the same defect: this becomes a check to perform in the project.
- Attach the analysed URL and evidence for the finding to the task.
- Specify the expected behaviour and pages to check in the code.
- Review the correction in context, then rerun the check on the same URL after deployment.
- Record unavailable measurements and human checks that remain necessary.
Read the report and choose whom to share it with
Anyone with the link can view the summary. Detailed findings require a verified account. Arrange that access for the person examining the evidence and keep the link among the people concerned.
For a first attempt, open the public example linked below, then choose a page from your site. Pick a finding whose evidence you can examine and whose post-fix check you can describe. Auditing a URL remains a starting point: a complete journey and checks requiring human judgement need additional work.
Sources and useful links
- Argos: start analysing a public URL, accessed 19 October 2026
- Argos: scoring method and measurement coverage
- Argos: educational example with fictional data
- Argos: report access and analysis limitations
- Argos: source code and documentation
- On this site: designing the Argos project
- In the Journal: understand a slow page before optimising it