A saved tutorial can look clean and still be unusable. The menu is gone, but so is the warning before a command. The code is present, but its indentation has collapsed. The screenshots appear, but the text no longer says which software version they describe. When keeping a tutorial for later reference, check whether you preserved the procedure, not just the paragraphs.

Start with a small preservation checklist

Before extracting anything, identify the parts the reader needs to carry out the task: prerequisites, supported versions, ordered steps, code or commands, expected output, screenshots that explain a setting, and recovery notes. Keep the original URL, author when stated, publication or revision date when available, and the date you saved the copy.

Do not silently turn the date of your saved copy into the tutorial's update date. A guide saved today may describe an older version. If the source does not state a version, record that uncertainty instead of guessing from a screenshot.

Choose the content boundary before removing clutter

Work from a page you own or are permitted to preserve. Compare the visible article with the saved result. A content container is a useful starting point, but a theme can place an author note, warning or download outside it. Conversely, a related-posts block can sit inside the article container.

Guangsuan's guide to extracting main content while excluding navigation and footer text explains structural selection and ways to recognise repeated template content. For tutorials, treat those methods as candidate filters and check what they remove. Repeated wording alone is not enough to discard a prerequisite that genuinely applies to every lesson in a series.

Inspect call-out boxes individually. An advertisement can go; an explanation of which operating system a command applies to needs to remain with that command. If you use automated rules, review both the kept text and the removed text for a sample before applying the rules to the rest of the collection.

Give code its own check

Flattening everything into plain paragraphs can damage code. MDN's documentation for the HTML preformatted text element explains that it preserves whitespace. Keep code and command blocks distinct from ordinary prose, preserving line breaks and meaningful indentation.

• Compare a multi-line block in the saved copy with the original, including indentation and blank lines where they matter.
• Make sure characters such as angle brackets have not disappeared because they were interpreted as HTML tags.
• Check that a displayed line number has not become part of the command.
• Keep the language or shell label, and distinguish user input from expected output.
• Check copied text in a plain-text editor. A screenshot that looks correct does not prove that its text can be copied correctly.

Text preservation is not execution testing. If you have not run the example in the stated environment, label it as preserved but untested. A command copied perfectly from an old guide can still be wrong for current software.

Check the dependencies that a text file cannot hold

Open image references and download links from the saved copy. Relative links can resolve differently after a file moves. A guide that says “use the setting shown below” loses meaning if the image was not saved or cannot be opened.

Record any external dependency you have not preserved, such as a downloadable sample project, a video or an interactive demonstration. If offline use is the goal, test the copy offline rather than assuming a browser's successful online display proves it is self-contained. Do not bypass access restrictions to fill a missing dependency.

Accept, flag or re-extract

Finish by walking through one representative procedure from start to finish as a reader. Check that the prerequisites are present, the steps remain in order, each code block has its context, and the expected result is still explained. You can do this reading check without executing commands.

Accept the copy when the required material is intact. Flag it when a known image, version detail or download is missing. Re-extract when the content boundary or formatting has damaged the procedure. Keep preservation status separate from technical validation status, so a future reader knows whether the guide was merely captured or actually tested.