Back to the trend of the day: Facebook's owner launches an AI assistant that remembers you, mostly for paying users
Poly Rating · from the trend of the day

AI That Remembers You: 8 Assistants Ranked (2026)

Meta says Muse Wonder remembers best, 85.6 against Claude Opus 5.5's 81.5, yet Meta ran every one of those scores itself, and its help page offers no way to switch memory off.

  • No. 1: ChatGPT memory
  • 8 places

01Final list

The leading AIs each built their own list of ten independently. Here is one result that is easy to check: every place has a reason, a source, and the list of AIs that included it.

Order: from most to least AI agreement.

A bright badge means this AI put the entry on its list. A dim badge means it did not.

  1. 1

    ChatGPT memory

    On by default, with two layers. Turning chat-history reference off schedules deletion within 30 days, but a saved memory goes only when you also delete its source chat.

    Included it: Claude, ChatGPT.

    In their lists: Claude: 3 · ChatGPT: 1

    What each AI noted 2
    • Claudeplace 3

      Memory is on by default and has two layers, but turning it off does not erase anything, and a saved memory is gone only once you also delete the chat it came from.

      Source ↗
    • ChatGPTplace 1

      It offers saved-memory controls and chat-history reference, but turning history reference off schedules remembered information for deletion within 30 days; source chats need separate deletion.

      Source ↗
    2 AIs put it here
    Source ↗
  2. 2

    Google Gemini memory

    The only one with a set auto-delete window (3, 18 or 36 months). But it can draw on Gmail, Photos and Search, and part of its automatic memory isn't shown.

    Included it: Claude, ChatGPT.

    In their lists: Claude: 4 · ChatGPT: 2

    What each AI noted 2
    • Claudeplace 4

      Its Personal Intelligence can draw on Gmail, Photos, Search and YouTube history, so memory sits inside the Google account, and part of the automatic layer isn't visible to the user.

      Source ↗
    • ChatGPTplace 2

      Google documents a 3-, 18-, or 36-month activity window, and turning activity off still allows conversations to be kept for up to 72 hours.

      Source ↗
    2 AIs put it here
    Source ↗
  3. 3

    Claude memory

    Shows every memory entry, with two separate on/off switches plus incognito. Its 81.5 memory score comes from Meta's run, and the controls are documented in one comparison blog.

    Included it: Claude, ChatGPT.

    In their lists: Claude: 1 · ChatGPT: 3

    What each AI noted 2
    • Claudeplace 1

      Its panel shows every memory entry, with two independent on/off toggles plus incognito, and Opus 5.5's vendor-reported MemoryBench 81.5 comes from Meta's run, not an outside test.

      Source ↗
    • ChatGPTplace 3

      Anthropic supports a memory feature on paid plans, but the benchmark and detailed current control settings were not verified from an opened primary help page in this research pass.

    2 AIs put it here
    Source ↗
  4. More about the top three ↓
  5. 4

    Microsoft Copilot memory

    Memory controls are documented mainly for Microsoft 365 work accounts, where admins hold the settings. Rollout dates and the consumer price conflict across sources.

    Included it: Claude, ChatGPT.

    In their lists: Claude: 8 · ChatGPT: 5

    What each AI noted 2
    • Claudeplace 8

      Its memory controls are documented mainly for Microsoft 365 work accounts, where admins hold the settings, and the consumer price and memory rules are not confirmed.

      Source ↗
    • ChatGPTplace 5

      A consumer assistant worth comparing, but no compatible published memory score or complete control/retention evidence was verified here.

      Source ↗
    2 AIs put it here
    Source ↗
  6. 5

    Perplexity memory

    Memories can be deleted one by one or all at once, and switching memory off stops new saves, at about $20 a month. There is no memory score and no auto-delete window.

    Included it: Claude, ChatGPT.

    In their lists: Claude: 6 · ChatGPT: 6

    What each AI noted 2
    • Claudeplace 6

      Memory saves mostly what you repeat and can be deleted entry by entry or wiped in one go, at the same $20 monthly price as Claude Pro and ChatGPT Plus.

      Source ↗
    • ChatGPTplace 6

      Its search-first product context makes data scope a key question, but a comparable public memory score and controls were not verified in this pass.

      Source ↗
    2 AIs put it here
    Source ↗
  7. 6

    Grok memory

    On by default and managed under Data Controls, but one comparison calls viewing and editing limited, and it isn't offered in parts of the EU and UK.

    Included it: Claude, ChatGPT.

    In their lists: Claude: 7 · ChatGPT: 7

    What each AI noted 2
    • Claudeplace 7

      Memory is on by default and managed under Data Controls, but it is not available in parts of the EU and UK, and one comparison calls its visibility and editing limited.

      Source ↗
    • ChatGPTplace 7

      No comparable public memory benchmark or verified retention/control window surfaced in the primary sources opened here.

    2 AIs put it here
    Source ↗
  8. 7

    Meta AI Muse memory

    Claims the best memory score at 85.6, measured only by Meta. No off switch is documented, Meta says Muse 'may still remember' deleted items, and Wonder Full costs $50/month.

    Included it: Claude, ChatGPT.

    In their lists: Claude: 5 · ChatGPT: 4

    What each AI noted 2
    • Claudeplace 5

      Meta claims the best memory score at 85.6, but its help page documents no way to switch memory off, and says Muse 'may still remember' what you delete.

    • ChatGPTplace 4

      Meta’s launch claims nearly identical MemoryBench scores for Wonder and Wonder Full—85.6 and 85.5—yet independent repetition of that headline test was not verified.

      Source ↗
    2 AIs put it here
    Source ↗
  9. 8

    Mistral Le Chat memory

    Its own help pages document an opt-out, edit, delete, 'Clear all', export and incognito. The free plan keeps up to 500 memories, but there is no published memory score.

    Included it: Claude. Did not include: ChatGPT.

    In their lists: Claude: 2

    What each AI noted 1
    • Claudeplace 2

      It is the one assistant here that documents an opt-out of memory, edit and delete from a Memories tab, 'Clear all', export, import and incognito chats, and its Free plan keeps up to 500 memories.

      Source ↗
    1 AIs put it here
    Source ↗

02More about the top three

The ranking is above. Here is more about the three places at the top: what they are, why they are here, and what is worth knowing.

The three are not measured the same way: ChatGPT and Gemini are judged on deletion rules (30 days; 3, 18 or 36 months), while Claude has a vendor-reported score of 81.5 from Meta's own run.

1ChatGPT memory

No. 1 — ChatGPT memory. On the lists of 2 of 2 AIs; median place there: 2. The AIs ranked it first because its two memory layers come with clear deletion rules, though a saved memory needs extra steps to remove.

Did you know

  • Deleting a chat from the sidebar doesn't free up saved memory. Chat history and saved memories are stored separately. You have to remove memory entries in Personalization settings, or ask ChatGPT to forget them.

    Source ↗ ✓ confirmed by several AIs
What it is
OpenAI's chatbot memory, with saved memories plus insights drawn from past chats.
Its edge
Gives the clearest deletion rule: turning off history reference schedules deletion within 30 days.
Who it suits
Suits people who want rich memory and will manage it; not people who want one-click erasing.
  • Memory controls and scoreDeletion within 30 days
Why it is here

  • Turning off chat-history reference schedules deletion of that information within 30 days. Saved memories have to be deleted separately, along with the chat where they began. — (voice only)

    Source ↗

  • ChatGPT uses two memory layers: saved memories and insights from past chats. That gives it more than a single list of notes to draw on. — (voice only)

    Source ↗
Did you know

  • ChatGPT's saved-memory list has a ceiling, but OpenAI doesn't publish the number. One guide estimates roughly 100 to 200 memories, and says that figure is its own estimate, not OpenAI's.

    Source ↗
History

  • In April 2025, ChatGPT added a background system that could draw on past chats, after its first memory feature depended on users asking it to save details. OpenAI described the earlier notes as prone to going stale or becoming irrelevant.

    Source ↗
Keep in mind

  • Main complaint: a Plus user reports no warning when memory is full, so new entries can be lost. Other users say saved memories vanish, and some still report missing data.

    Source ↗
What people say

  • "I might mistakenly believe that my information has been successfully saved, only to later discover that it was lost." — maverick.liou (OpenAI community forum)

    Source ↗
Compare the top three ↓

2Google Gemini memory

No. 2 — Google Gemini memory. On the lists of 2 of 2 AIs; median place there: 3. The AIs placed it second for its set auto-delete windows, held back because it can draw on Gmail, Photos and Search and hides part of its memory.

Did you know

  • Turning Keep Activity off doesn't erase chats at once. Google says conversations are still saved with your account for up to 72 hours, and they won't show in your activity list.

    Source ↗
What it is
Google's Gemini memory, tied to your Google account and Gemini Apps Activity.
Its edge
Only one of the three with a set auto-delete window you choose: 3, 18 or 36 months.
Who it suits
Suits people already living in Google apps; not people who want memory kept apart from their data.
  • Memory controls and score3, 18 or 36 months
Why it is here

  • Gemini Apps Activity is set to auto-delete after 18 months by default; users can choose 3 months or 36 months instead. — (voice only)

    Source ↗ ✓ confirmed by several AIs

  • With Personal Intelligence, Gemini can connect information across Gmail, Photos, Search and YouTube, so its memory can sit beside several kinds of personal data. — (voice only)

    Source ↗
History

  • In March 2025, Gemini’s personalization feature began using Google Search history to tailor replies, with the user’s permission. Google said Photos and YouTube access would follow, starting a shift from saved preferences toward context from connected apps.

    Source ↗
Keep in mind

  • Google's privacy policy lets chats be stored up to three years or used for AI training. Turning off past-chat memory doesn't undo chats already used for training.

    Source ↗

  • At launch, the memory feature was limited. It started with the 2.5 Pro model, wasn't available in the UK, and wasn't offered to users under 18. Check what your account gets today.

    Source ↗
What people say

  • "Google is introducing a new ability for Gemini: the AI can now remember your past chats by default." — Henderson Jayden Harper (MakeTechEasier)

    Source ↗
Where to go further

  • Look at the auto-delete section first. It lets you keep activity for 3, 18 or 36 months, or turn auto-delete off. The default is 18 months. Pick the window you want before you use Gemini much.

    Source ↗
Compare the top three ↓

3Claude memory

No. 3 — Claude memory. On the lists of 2 of 2 AIs; median place there: 2. The AIs put it third because its controls are clear and visible, but its score is Meta's own run and the controls rest on one blog.

Did you know

  • Claude's memory can move between assistants. Maginative reports you can pull your memory data from ChatGPT or Gemini and move it into Claude, so switching doesn't mean starting from zero.

    Source ↗
What it is
Anthropic's Claude memory, with a panel showing entries, two switches and incognito chats.
Its edge
Shows every memory entry and gives two separate on/off switches plus incognito.
Who it suits
Suits people who want to see and edit what is stored; not people who need proof from independent tests.
  • Memory controls and score81.5 memory score

Wikidata

Why it is here

  • Claude’s memory controls include separate settings for searching chats and managing memory; users can also export their memory for backup or migration. — (voice only)

    Source ↗
Did you know

  • Each Claude Project has its own separate memory. For example, product launch planning stays apart from client work. Incognito chats are kept out of memory entirely.

    Source ↗
History

  • In September 2025, Anthropic first introduced Claude memory to Team and Enterprise users, after safety tests raised concerns such as memory reinforcing harmful patterns. It adjusted the feature before launch; Pro and Max users got it in October.

    Source ↗
Keep in mind

  • Claude doesn't keep full conversations. It writes summaries of them. Incognito chats aren't saved to memory or history, so anything you want remembered must be in a normal chat.

    Source ↗

  • Plan gap: one guide says Team and Enterprise accounts still use the older summary-style memory until the newer version reaches them. What you see can depend on your plan.

    Source ↗
What people say

  • "Claude is now building a picture of who you are across conversations, quietly, in the background." — Syracuse University IT

    Source ↗
Where to go further

  • Read how to view, edit and delete single memories, and how incognito mode works. The menu names differ between versions of this help page, so check your app's Settings for the exact labels before you start.

    Source ↗
Claude put it first. What they saw in it

  • Claude: Its panel shows every memory entry, with two independent on/off toggles plus incognito, and Opus 5.5's vendor-reported MemoryBench 81.5 comes from Meta's run, not an outside test.

Video: Every Claude Code Concept Explained for Normal People · Simon Scrapes

Compare the top three ↓

The top three side by side

ChatGPT memoryGoogle Gemini memoryClaude memory
In the AIs' lists2 / 22 / 22 / 2
Memory controls and scoreDeletion within 30 days3, 18 or 36 months81.5 memory score
Put first byChatGPT—Claude

Places with the same number of AIs are ordered by their average place in the AIs' own lists.

More ratings

03Worth knowing

What the AIs noticed while building the list.

Meta's lead (85.6 against 81.5 for Opus 5.5) comes from a Meta-built, LLM-judged test in which Meta ran the rivals itself. No independent memory test of any consumer assistant was found.

Noticed by:
Included it: Claude, ChatGPT.
Source ↗

Switching memory off is not erasing it. ChatGPT keeps a saved memory until you also delete its source chat, Le Chat keeps memories until you use 'Clear all', and Meta says Muse may still remember deleted items.

Noticed by:
Included it: Claude, ChatGPT.
Source ↗

Gemini's memory sits closest to the rest of your digital life: opt-in links to Gmail, Photos, Search and YouTube, and saved activity can include audio, Live video and location.

Noticed by:
Included it: Claude, ChatGPT.
Source ↗

The price of full memory ranges from free (Le Chat, up to 500 memories) to $50 for Meta's Wonder Full and $100–$200 for the top plans at OpenAI, Anthropic and Google.

Noticed by:
Included it: Claude. Did not include: ChatGPT.
Source ↗

LongMemEval splits memory into five abilities: extraction, multi-session reasoning, temporal reasoning, knowledge updates and abstention. One vendor score can't stand for all of them.

Noticed by:
Included it: ChatGPT. Did not include: Claude.
Source ↗

04Questions people ask

Which AI remembers you?

ChatGPT, Claude, Gemini, Grok, Perplexity, Le Chat and Meta's Muse all keep memory across chats. Meta claims the top memory score, 85.6, but only Meta has measured it.

Source ↗
ChatGPT vs Claude memory: which is better?

Both are on by default and cost $20 a month at the standard tier. Meta's run puts Opus 5.5 at 81.5 and GPT-6 Astra at 81.4, too close to separate, while Claude shows every stored entry.

Source ↗
ChatGPT vs Claude memory Reddit: why do people disagree?

No like-for-like independent memory test exists, so opinions rest on anecdotes. The only shared score, 81.5 against 81.4, comes from Meta's own run.

Source ↗
Claude AI vs ChatGPT memory: how do I delete it?

In ChatGPT, delete the saved memory and the chat it came from. Turning history reference off schedules deletion within 30 days. In Claude, delete entries in the memory panel, or pause or reset memory.

Source ↗

05How we counted

In plain words, no formulas.

1The question

Top 10 AI Assistants That Remember You, Ranked by How Much Control You Keep Over What They Remember

2Why this topic

Readers can see that the assistant that remembers best is not always the one that lets them see and delete what it remembers, a trade-off no single table shows.

3Where it applies

global, consumer AI assistants with memory across conversations

4What we looked at

  • Memory strength from a published memory benchmark, marked as vendor-reported or independent
  • How much of the stored memory the user can view, edit, delete and switch off, and whether there is an auto-delete window
  • Whether memory is on by default
  • What other personal data the memory sits beside (email, search history, photos, social posts, messages)
  • Monthly price of the tier that gets full memory
  • Independent evidence of real use: human preference votes and live agent task outcomes, kept separate from vendor claims
  • MemoryBench (Meta, vendor-reported) — A Meta-built, LLM-judged test of recalling facts across long conversations. Meta ran the rival models itself on a 'best-effort' basis.
  • Vendor memory documentation (OpenAI, Google, Meta, Mistral) — Defaults, view, edit and delete controls, off switches and retention windows.

5Who ranked

The leading AIs, each building its own list of ten independently and naming a source for every place: Claude, ChatGPT.

6How the places were set

The same entry written differently by different AIs was counted as one. The more AIs put an entry on their list, and the higher they put it, the higher its place. "AI agreement" shows how many of the AIs had it on their own list. Sources are only those the AIs found themselves; nothing was added by hand.

06Each AI's own list

🇺🇸 Claude8 places · 4 notes

The same question, answered by Claude alone. The final list above combines all the AIs.

  1. 1
    Claude (Anthropic)

    Its panel shows every memory entry, with two independent on/off toggles plus incognito, and Opus 5.5's vendor-reported MemoryBench 81.5 comes from Meta's run, not an outside test.

    Source ↗
  2. 2
    Mistral Le Chat (Mistral AI)

    It is the one assistant here that documents an opt-out of memory, edit and delete from a Memories tab, 'Clear all', export, import and incognito chats, and its Free plan keeps up to 500 memories.

    Source ↗
  3. 3
    ChatGPT (OpenAI)

    Memory is on by default and has two layers, but turning it off does not erase anything, and a saved memory is gone only once you also delete the chat it came from.

    Source ↗
  4. 4
    Gemini (Google)

    Its Personal Intelligence can draw on Gmail, Photos, Search and YouTube history, so memory sits inside the Google account, and part of the automatic layer isn't visible to the user.

    Source ↗
  5. 5
    Meta Muse with Wonder (Meta, 2026)

    Meta claims the best memory score at 85.6, but its help page documents no way to switch memory off, and says Muse 'may still remember' what you delete.

  6. 6
    Perplexity

    Memory saves mostly what you repeat and can be deleted entry by entry or wiped in one go, at the same $20 monthly price as Claude Pro and ChatGPT Plus.

    Source ↗
  7. 7
    Grok (xAI)

    Memory is on by default and managed under Data Controls, but it is not available in parts of the EU and UK, and one comparison calls its visibility and editing limited.

    Source ↗
  8. 8
    Microsoft Copilot

    Its memory controls are documented mainly for Microsoft 365 work accounts, where admins hold the settings, and the consumer price and memory rules are not confirmed.

    Source ↗

What Claude noticed

Meta's MemoryBench lead (85.6 vs 81.5 for Opus 5.5) comes from a Meta-built test with an LLM judge and Meta-run competitor scores. I found no independent MemoryBench result, and no independent memory test of any consumer assistant.

Source ↗

Turning memory off is not the same as erasing it: ChatGPT keeps saved memories until you delete them and the chat they came from, Le Chat keeps stored memories until 'Clear all' is used, and Meta says Muse may still remember what you deleted.

Source ↗

Gemini's memory sits closest to the rest of your digital life: opt-in links to Gmail, Photos, Search and YouTube. Google says it trains on prompts and responses after filtering personal information, not on the connected data itself.

Source ↗

The price of the strongest memory tier varies about tenfold across the list: free (Le Chat, up to 500 memories) or $14.99–$20 for the mainstream tiers, against $50 for Meta's Wonder Full and $100–$200 for top plans at OpenAI, Anthropic and Google.

Source ↗
🇺🇸 ChatGPT10 places · 4 notes

The same question, answered by ChatGPT alone. The final list above combines all the AIs.

  1. 1
    ChatGPT memory

    It offers saved-memory controls and chat-history reference, but turning history reference off schedules remembered information for deletion within 30 days; source chats need separate deletion.

    Source ↗
  2. 2
    Google Gemini memory

    Google documents a 3-, 18-, or 36-month activity window, and turning activity off still allows conversations to be kept for up to 72 hours.

    Source ↗
  3. 3
    Claude memory

    Anthropic supports a memory feature on paid plans, but the benchmark and detailed current control settings were not verified from an opened primary help page in this research pass.

  4. 4
    Meta AI / Muse memory

    Meta’s launch claims nearly identical MemoryBench scores for Wonder and Wonder Full—85.6 and 85.5—yet independent repetition of that headline test was not verified.

    Source ↗
  5. 5
    Microsoft Copilot memory

    A consumer assistant worth comparing, but no compatible published memory score or complete control/retention evidence was verified here.

    Source ↗
  6. 6
    Perplexity memory

    Its search-first product context makes data scope a key question, but a comparable public memory score and controls were not verified in this pass.

    Source ↗
  7. 7
    Grok memory

    No comparable public memory benchmark or verified retention/control window surfaced in the primary sources opened here.

  8. 8
    Pi memory

    Pi has a personal-assistant framing, but no comparable published memory benchmark or fully documented controls were verified here.

  9. 9
    Character.AI memory

    Long-running character conversations make continuity relevant, but continuity is not the same as user-auditable personal memory.

  10. 10
    Replika memory

    It is built around a continuing relationship with a user, but no comparable independent memory score or control/retention comparison was verified here.

What ChatGPT noticed

LongMemEval’s authors define five abilities—information extraction, multi-session reasoning, temporal reasoning, knowledge updates, and abstention—so a single vendor memory score should not be read as a complete measure of remembering a person.

Source ↗

Gemini’s documented data scope can extend beyond chat text: Google lists audio, Live videos and screenshares, visited-site information, product usage, and location data among possible saved activity.

Source ↗

Meta’s claimed MemoryBench scores for Muse Wonder and Wonder Full differ by only 0.1 point, while AA reports a five-point separation on its Coding Agents Index; the tests measure different capabilities.

Source ↗

A study of 1,999 U.S. assistant users reported ChatGPT as the primary assistant for 58% and Gemini for 25%; market use is not evidence of memory quality, but it is a useful adoption signal.

Source ↗