> ## Documentation Index > Fetch the complete documentation index at: https://docs.zued.ai/llms.txt > Use this file to discover all available pages before exploring further. # Export your data > Download every AI answer from a snapshot, with its citations, sources and searches, as CSV files in one zip. The **data export** hands you the raw answers behind your numbers: every answer the AI engines gave in one [snapshot](/platform/snapshots), with its citations, sources and the searches the engines ran. You get a zip of CSV files and a README that explains every column, ready for a spreadsheet, your own BI tool or an archive outside zued. Every member of a project can start an export and download any export in its list, viewers included. See [Team members](/platform/projects#team-members). ## Starting an export **Export data** sits in the bar at the top of the Overview, next to Search. It becomes available once your first snapshot has finished. 1. Pick the **Snapshot**. It starts on the snapshot selected in the bar. **Latest** exports the newest finished snapshot; earlier weeks are in the same list. 2. Pick the **Answers**: * **One answer per question**: the answer zued shows for each prompt and engine. * **All runs**: every time zued asked, up to 10 times per prompt and engine. This is the data behind the percentages. 3. Click **Start export**. The export builds in the background, and the **Recent exports** list in the same dialog shows its progress. Each row names the snapshot's date and, once ready, its answer count and file size. One export runs per project at a time, and a snapshot that is still running can be exported once it has finished. ## Getting the file * **Automatic download.** The tab you started the export in downloads the zip on its own when it is ready, whichever view of that project you are on, and shows a toast with a **Download** action. * **Download button.** Every ready export keeps a **Download** button in the Recent exports list for 7 days. Use it after closing the tab, on another device, for an export a teammate started, or when your browser blocked the automatic download. * **Email.** The person who started the export gets an email when it is ready. The email links into zued, where you sign in to download. It holds no file and no download link, so forwarding it gives nobody the data. Every download, the automatic one included, uses a fresh link that stops working after 1 hour. ## What's in the zip | File | One row per | | - | - | | `answers.csv` | answer to a prompt, with the full answer text, the engine, the run and when it was collected | | `citations.csv` | link an answer cites as a source | | `sources.csv` | page in the engine's search results for an answer | | `engine_queries.csv` | search the engine ran, follow-up it suggested, or question it asked back | | `prompts.csv` | prompt answered in the snapshot, with its topic, persona and page | | `pages.csv` | page those prompts belong to | | `fact_check_answers.csv` | answer to a fact-check prompt, with the page's accuracy verdict for the snapshot | | `fact_check_claims.csv` | checked claim and engine, with what the engine said and its verdict | | `README.md` | the data dictionary: every column, the row count of each file, and notes on this snapshot | | `manifest.json` | the README's facts in machine-readable form, including which content is AI-generated | Rows carry IDs, so you can join the files. A prompt keeps one ID in every snapshot, so you can line up exports from different weeks. `answers.csv` and `fact_check_answers.csv` repeat the prompt text and page next to each answer, so they read without a lookup. ### The counted column `answers.csv` and `prompts.csv` carry a `counted` column. It is true when zued's numbers for that snapshot count the answer or prompt. The files keep every answer the snapshot holds, and the ones zued does not count are marked false. Filter `counted` to true before you pivot. zued's default numbers also leave out prompts that name your own brand (see [Brand exclusion](/platform/prompts#brand-exclusion)). `counted` does not mark those. ### Fact-check answers have their own files [Fact-check prompts](/platform/prompts#managing-prompts-after-approval) measure whether engines state your facts correctly. Their answers sit in `fact_check_answers.csv` and `fact_check_claims.csv`, so a pivot of `answers.csv` only counts the prompts behind your visibility numbers. Their citations, sources and searches are left out of the other files for the same reason, and the README counts what was left out. ### Opening the files in Excel Open the CSVs through **Data, From Text/CSV**. A double-click can make Excel read answers that start with `=`, `+`, `-` or `@` as formulas. The README also flags any file with more rows, or any cell with more characters, than Excel can hold. ## What the export does not contain * **zued's computed metrics**, such as [Appearance Rate](/metrics/appearance-rate) or [Share of Voice](/metrics/share-of-voice). Those stay in the app. The export holds what the engines said. * **Your page content.** The export covers what the engines answered. `pages.csv` lists your pages by URL and title. * **Answers past the retention window.** A snapshot whose individual answers were pruned (see [Snapshot history](/platform/snapshots#snapshot-history)) can no longer be exported. Keep the zip if you need a snapshot's answers later. ## How long the file is kept The zip is deleted 7 days after it is ready. When a project is permanently deleted (see [Archiving a project](/platform/projects#archiving-a-project)), its exports are deleted with it. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.