Research Guides
Match a Research Download Label to the Browser Action
A research page promises “Download the worksheet,” but selecting the link opens a new browser document. The file may be correct while the delivery experience is wrong. Treat the action a reader expects, the response instructions and the observed browser behavior as separate evidence.
This guide is a delivery review for publishers of research downloads. It is not a file-integrity test, an indexing guarantee or a claim that Colquhoun.AI provides a particular export feature.
1. State the promised action
Before inspecting headers, write down what the page promises. Does “View report” mean opening a readable document? Does “Download worksheet” mean saving a file for later editing? A reviewer cannot resolve a disagreement between the interface and the response if nobody has specified the intended experience.
MDN's Content-Disposition reference explains the response header's inline and attachment dispositions. The attachment form indicates download handling and can suggest a filename. Browser behavior also interacts with the HTML download attribute; when honored, that attribute can override inline behavior. Filename parameters can influence the suggested saved name, and browsers may transform names to meet filesystem requirements.
2. Review a fictional worksheet link
Imagine a fictional article with two actions: “Preview the methodology” and “Download the blank worksheet.” Both currently point to the same PDF route, which supplies an inline disposition. In the tested browser, both actions open the PDF viewer.
The reviewer records the article labels, target addresses, final response headers and resulting behavior. The finding is not that PDFs always need attachment handling. It is that the download action and the tested delivery path disagree. The preview action may already be appropriate.
A proposed repair gives the actions clearly defined purposes. The publisher might retain an inline preview and provide a deliberately tested download route. If the browser or platform cannot reliably offer the promised action, clearer instructions may be more honest than a misleading button.
3. Keep a compact delivery matrix
- Entry point: page address and exact visible action label.
- Target: link destination and final address after redirects.
- Response: status, media type and disposition header.
- Link behavior: relevant download attribute and navigation target.
- Environment: browser, device and settings that affect opening files.
- Observed outcome: viewer, save prompt, automatic save, error or other result.
- Saved name: actual name offered or produced, if a save occurs.
Use one row per action and environment. A desktop result should not silently stand in for a phone, where saving and finding the file can be a different journey. Keep the original values in the evidence rather than replacing them with a general “download works” label.
4. Retest the whole reader journey
Use a harmless public fixture or an authorized test document. Start at the article rather than pasting the destination into the address bar, because that would skip the link behavior being reviewed. Confirm that a reader can identify the action, complete it and find the resulting file where the tested platform puts it.
Check whether the proposed saved name is useful outside the browser. A generic name can make several research worksheets hard to distinguish. Avoid inserting private customer names or confidential project details into filenames just to make them memorable.
Do not overwrite a reader's existing file during a test. Keep download verification separate from opening untrusted active content, and do not weaken browser security controls to make a fixture appear to pass.
5. Close only the delivery finding
A good completion note names the tested actions and environments, the intended behavior and the observed result after the repair. If only the response headers were inspected, say so: that is not a browser interaction test.
Successful delivery does not establish that a document is current, accessible, complete or factually correct. Those remain separate checks. A narrow delivery record helps another publisher reproduce the experience without turning a working save action into an unsupported claim about the research itself.