How to read conversation JSON without being a programmer

JSON is a text format for structured records. If your conversation export includes a JSON file, you do not need to understand every field to make a useful first inspection. You do need to preserve the original and avoid tools that quietly transfer its contents somewhere else. Begin with a local text editor and a small, specific question.

Learn the few shapes that matter

Curly braces usually enclose an object: a set of named fields and values. Square brackets enclose a list. Quoted text is a string, and numbers or true/false values may appear without quotes. A field called “title” can help you identify a record, but field names are not a universal contract across providers or export versions.

A simple invented example is an object with a title, a date and a messages list. The messages may themselves be objects with a speaker and text. Real exports can be much more nested. Treat this example as a reading aid, not as the format that every importer expects.

Work on a copy

Duplicate the file into a private working folder. Open it in a local editor that can handle its size, preferably without cloud collaboration enabled. Large archives can exceed an editor’s comfortable memory limits. If the application freezes, close it and try a more suitable local viewer rather than repeatedly opening additional copies.

Do not save changes over the source archive. Automatic formatting may alter whitespace harmlessly, but accidental edits can break the structure. Keeping the original allows you to distinguish a damaged working file from a provider export problem and makes it possible to restart your inspection.

Search for a known conversation

Choose a project title or an unusual phrase you remember using. Search locally and inspect the nearby fields. Note whether the match appears in your text, an assistant response or metadata. The same phrase may occur several times, and a search hit alone does not show who said it.

Then inspect the surrounding record. Identify any speaker labels and available dates. If text appears as several segments, read them together before extracting a conclusion. When the structure is unclear, record the conversation title and return to a readable viewer instead of assuming that nearby text belongs to the same message.

Keep conversion decisions narrow

Online “JSON beautifier” and converter tools are convenient, but pasting private conversations into one is a disclosure. A claim that a tool runs locally should be checked against its actual behavior and policy. Prefer existing local functionality when your goal is simply searching or reading.

If you need a spreadsheet, extract a small set of non-sensitive fields such as neutral project labels and review dates. Do not flatten an entire archive simply because the software can. A spreadsheet can lose speaker distinctions, branch structure and context, so retain source references and describe what you omitted.

Know when to stop inspecting

Your goal might be finding one checklist or deciding whether an import contains recent history. Once that question is answered, further parsing may add little value. Avoid interpreting hidden, unsupported or unfamiliar fields as meaningful personal evidence.

Persir is not a conversation-JSON viewer and no longer accepts full ChatGPT or Claude exports. Its import accepts only the small summary response requested by the Persir prompt, including practical text or Markdown wrappers around that JSON. Keep general archive inspection local, and use the prompt flow when you want to add confirmed personal statements to your Persir profile.