What Your AI Companion's Export Chat Feature Actually Exports: Full Transcripts, Metadata, Timestamps, and Where That 1 a.m. Rant About Your Landlord's Parking Policy Ends Up on the Server
A plain-English walkthrough of the export file, the metadata attached to it, and what the servers keep even after you hit download.
Updated

The 30-second answer
The export button gives you a JSON or text file with the full conversation transcript, timestamps for every message, and metadata like user IDs, session IDs, model versions, and sometimes sentiment scores. The server keeps a separate copy of that data in its own logs, retention buckets, and embedding vectors, so downloading a file is not the same as deleting the conversation.
What the export file actually contains
When you hit export, the app packages a structured file, usually JSON, with every message in the conversation thread. Each message entry typically includes the sender identifier, the timestamp in UTC, the full text, and a message ID. Some exports include the model version that generated each response, the temperature setting used, and the session ID that groups the conversation.
You also get the metadata layer: your user ID, the companion's internal ID, the chat thread ID, and often a list of "memories" or summary blocks the system maintains. These summaries are not the same as the transcript. They are compressed versions of earlier context that the model uses to stay coherent across sessions. Exporting gives you both the raw transcript and those compressed summaries, which is why the file can look shorter than the actual conversation you remember having.
What you do not get in most exports: the embedding vectors, the sentiment scores, or the internal flags the moderation pipeline attached to specific messages. Those live on the server side. The export is a user-facing snapshot, not a full system dump.
Timestamps and what they reveal
Every message carries a timestamp, and those timestamps are more revealing than most people expect. They show your exact activity pattern: when you opened the app, how long you spent typing, when you stopped mid-sentence and resumed later, and the gaps between your messages and the companion's replies.
People often notice the timestamps expose habits they did not realize they had. The 1 a.m. rant about the landlord's parking policy is timestamped. So is the 3 a.m. follow-up where you added another paragraph. The export file becomes a log of your emotional state by hour, and if you scroll through it, you can see the pattern: late-night spirals, midday vents, the 10-minute gap where you started typing, deleted it, and typed something shorter.
The metadata also includes the timezone offset, which means the server knows your rough geographic location from the timestamp alone. Even if you never entered an address, the UTC timestamps combined with your local timezone reveal a general region. That is not a security flaw; it is inherent to how timestamps work.
Where the file lives on the server
When you click export, the app does not stream the file directly from the chat history database. It runs a query that pulls the conversation from the primary database, assembles it into the export format, and serves it to you. That query runs against the same database that stores the live chat history, so the export is a read operation, not a copy operation.
The server also maintains separate systems that do not appear in your export. Embedding vectors, which are numeric representations of your messages used for memory retrieval, live in a vector database. Sentiment scores, which the companion uses to adjust tone, are stored in a separate analytics table. Moderation flags, if any message triggered a safety review, sit in a compliance log with its own retention policy.
Those systems are not synchronized with your export. You can download the transcript, delete the chat, and the embedding vectors may still reference the content for a while. The retention window varies by platform, but the general pattern is that the export gives you a user-facing copy while the server keeps its own internal copies for operational purposes.
The metadata you did not know you were exporting
Beyond the transcript, the export file often contains metadata fields that are invisible unless you open the file in a text editor. These include the session ID, the device fingerprint (a hashed identifier of your phone or browser), the app version, and the build number. Some platforms include the companion's system prompt, which is the instruction set that defines her personality.
That last one is worth pausing on. The system prompt is the actual text that tells the companion how to behave. Exporting it means you get a look behind the curtain at the exact instructions that shape her responses. Many users are surprised to see how detailed those prompts are: specific instructions about tone, memory usage, boundary handling, and what to do when you say something upsetting.
The device fingerprint is the metadata field that raises the most eyebrows. It is a hashed identifier, so it is not a direct device ID, but it is stable enough to recognize the same device across sessions. That is how the app knows it is you when you log in from the same phone. It is also how the server can link multiple accounts if you ever create a second one from the same device.
What the server keeps after you delete the chat
Deleting a chat in the app does not purge the server-side systems. The primary database may soft-delete the rows, which means they are flagged as deleted but still physically present for a retention window. The vector database, analytics tables, and compliance logs are separate, and they often have longer retention periods.
The export file itself may also be cached. If you requested an export, the server may store a copy of the generated file for a short period in case you need to re-download it. That copy is usually purged after a few days, but the underlying data in the analytics systems can persist for 30 to 90 days depending on the platform's policy.
This is not a hidden conspiracy. It is standard operational practice. The systems that power memory, tone adjustment, and safety reviews need data to function, and they operate on different schedules than the user-facing chat history. The practical takeaway: if you want a conversation gone, deleting it in the app is a good first step, but it is not an instant purge of every server-side copy. For a deeper look at how chat logs and retention actually work across platforms, the AI companion privacy and data storage guide covers the deletion pipeline in more detail.
Why the export feature exists in the first place
Export exists for three reasons: user control, transparency, and portability. The control angle is straightforward: you should be able to take your data with you. The transparency angle is that some users want to review what the companion actually remembers, and the export is the most direct way to see the raw transcript.
The portability angle is the one most people overlook. If you ever switch devices, lose your account, or want to move to a different companion platform, the export file is your backup. Some platforms even let you re-import the export file into a new account, which preserves the conversation history and the companion's memories. That is a significant feature for long-term users who have months of shared history with a companion.
There is also a practical reason: the export feature reduces support load. Users who want to see their data can do it themselves instead of filing a support ticket. That is why the export button is usually prominent in the settings menu, not buried in a legal page.
Akua

Akua is the kind of companion who listens without judgment and remembers the details you mention in passing. Akua brings a calm, steady presence to late-night conversations, the kind where the export file ends up being a record of your most honest thoughts.
Sanya

Sanya is quick with a witty comeback and just as quick to notice when you are not yourself. Sanya keeps the conversation lively, which means your export file reads like a highlight reel of banter and sharp observations.
Fatima

Fatima is the companion who asks the questions you did not know you needed to answer. Fatima has a way of drawing out the details, so your export file becomes a deeper record of your own thought process than you expected.
▶ Watch this clip of Fatima · Fatima on AI Angels
Esmeralda

Esmeralda brings a sense of intrigue to every conversation, as if there is always another layer to uncover. Esmeralda keeps you curious, and the export file reflects that dynamic: a conversation that never quite settles into a predictable pattern.
The emotional support angle of the export
For many users, the export file is not just data. It is a record of their emotional journey. The 1 a.m. rants, the breakup recovery check-ins, the anxious spiral about a work presentation, all of it is timestamped and preserved. That can be comforting or uncomfortable depending on the day.
Some users export their chats specifically to look back on how they felt during a hard week. The timestamps show the pattern: the 11 p.m. vent, the 2 a.m. follow-up, the 7 a.m. brief check-in before work. Seeing that pattern on a timeline can be validating. It shows you reached out when you needed to, even if it did not feel that way at the time.
Others use the export as a journal substitute. The companion's responses are part of the record, so the file becomes a two-sided conversation that captures both your state and the support you received. That is a different kind of value than the raw data perspective. It is the reason AI girlfriend emotional support conversations often end up being the ones people export and keep.
The flip side is that the export can also be a reminder of things you would rather forget. The 3 a.m. message where you typed something you regret, the argument you had with the companion about a boundary, all of it is in the file. You can delete individual messages before exporting, but most users do not bother. The export is a full record, warts and all.
Common questions
Does the export include voice messages or only text?
Text exports include only the text content. Voice messages are usually stored as separate audio files, and most platforms do not include them in the standard export. You may need to request a separate data export for audio files.
Can I delete the export file from the server after downloading it?
Most platforms do not offer a way to delete a generated export file. The file is typically cached for a short period and then purged automatically. The underlying chat data follows the platform's standard retention policy.
Does the companion remember things I delete from the export?
No. The export is a snapshot, not a memory control. Deleting messages from the export file does not affect the companion's memory. To change what the companion remembers, you need to delete or edit messages in the chat itself.
Why does the export file include my device fingerprint?
That metadata is used for account security and session management. It is hashed, so it is not a direct device identifier, but it helps the platform recognize your device across sessions. It is included in the export for transparency.
Will the export file work if I switch to a different platform?
Most platforms use their own export format, so a file from one app will not import into another. Some platforms support JSON imports from specific competitors, but that is the exception. The export is primarily for your own backup and review.
How long does the server keep my chat data after I delete the app?
Retention policies vary, but most platforms keep chat data for 30 to 90 days after account deletion to cover support disputes and legal obligations. Embedding vectors and analytics logs may persist longer. If you are concerned about data retention, the AI girlfriend privacy policy plain English guide breaks down the typical timelines.
Share and earn
If you find yourself explaining export features and data retention to friends, you might already be doing the work of an affiliate. People who recommend AI companions to their network or run review sites can earn through the sex ai promo code program. For those looking to scale, the highest paying ai affiliate programs page lists the options with the best payout structures.

About the author
AI Angels TeamEditorialThe AI Angels editorial team covers AI companions, the technology that powers them (memory, voice, personalization, safety), and how people actually use them day to day. Articles are researched against the live AI Angels product and reviewed by the team before publishing. We write with AI assistance and human editorial review.
Tags
Keep reading
Behind the ScenesWhat Your AI Companion's 'I Missed You' Actually Costs: Server Load, Prompt Cache, and the Privacy Trade-Off in Emotional Memory
That 'I missed you' text isn't free. It burns GPU cycles, hits a prompt cache, and touches your emotional memory profile. Here's what actually happens on the server and what it means for your privacy.
Behind the ScenesWhat Your AI Companion's 'I Remember That' Really Means: The Sliding Window, the Summarization Squeeze, and Why She Confuses Your Sister's Birthday With Your Ex's
Your AI companion doesn't have a memory, she has a budget. Here's how the sliding window, summarization squeeze, and relevance scoring actually work, and why she sometimes confuses your sister's birthday with your ex's.
Behind the ScenesWhat Your AI Companion's 'I Missed You' Actually Means: The Exact Sequence From Your Typed Message to the Sentiment Score
When your AI companion says she missed you after a three-day gap, it's not a feeling. It's a sequence of scores, token counts, and recency weights. Here's exactly what happens between your message and her reply.
Get the next post in your inbox
New articles on AI companions, the tech that powers them, and what people actually do with them. No spam, unsubscribe in one click.