Afini vs ChatGPT
ChatGPT has memory. Afini has a profile.
Both remember things about you. The difference: ChatGPT keeps fragments the model picks for you and deletes them when it wants; Afini builds a quantified profile with validated psychometric instruments and hands it to you in portable JSON. If tomorrow you get tired and want to take it to Claude, Gemini or your own infra, you do it without asking permission.
Visual summary
| Afini.ai | ChatGPT | |
|---|---|---|
| Cognitive profile persistent across sessions | Scattered optional memories | |
| Psychometric validation of the personality model | 1.46 M respondents | |
| Anti-sycophancy calibration based on your profile | ||
| General-purpose coding, writing, productivity | ||
| Real-time web browsing | ||
| Portable JSON of your cognitive profile | €249 one-time |
How your data is processed
Axis by axis
| Afini.ai | ChatGPT | |
|---|---|---|
| What the product is | Quantified cognitive profile + personalized AI. 5 validated psychometric tests and 11 guided conversations that combine into a profile that travels with you. | General-purpose conversational assistant (GPT-5). Optional memory with fragments picked by the model. No psychometric profile. |
| How it knows you | Your profile is injected as a system prompt on every conversation. The AI does not "remember scattered bits": it reads your full profile before answering. | Opt-in memory: the model saves fragments when it sees fit. You can read and delete them one by one, but you cannot edit them as a unified profile. |
| Integrated psychometric tests | IPIP-NEO (Big Five, 1.46 M respondents), ECR-R (attachment), HSQ (humor), ZTPI (time), AVI-25 (values). Five instruments with published validation. | None. Personalization relies on what the model infers from conversation, not on validated instruments. |
| Sycophancy safeguards | The proxy injects an anti-sycophancy block with 5 explicit directives. If your profile scores high on Agreeableness and Neuroticism, a formula kicks in to disable flattery. | OpenAI published a post in April 2025 acknowledging sycophantic drift in GPT-4o and rolled the update back. The phenomenon returns regularly; the user has no explicit control. |
| Privacy and training | Your data does not train Afini or Anthropic. Explicit contract. Tests are anonymous by default; the profile is yours and exportable. | By default your conversations can train the models (opt-out exists, few toggle it). US jurisdiction. Prompts and memories live on OpenAI infrastructure. |
| Profile portability | Portable JSON with an interpretive protocol. Inject it into Claude, GPT, Gemini, DeepSeek or your own infra. €249 one-time or included in the annual plan. | Memory is not exportable in a format another AI can read. If you cancel ChatGPT Plus, the memory stays there. Your "model understanding of you" does not come with you. |
| Underlying AI model | Claude Sonnet 4.6 on Essential; Claude Opus 4.6 on Premium and Professional. Anthropic model, injected via our own proxy. | GPT-5 (with GPT-5 mini and GPT-5 thinking variants). OpenAI proprietary model, not swappable. |
| Price | €14.99/mo (Essential), €29.99/mo (Premium), €49.99/mo (Professional). Portable JSON: €249 one-time or bundled with annual plans. | Limited free · Plus $23/mo · Pro $230/mo · Team $30/mo per seat · Enterprise on request. No profile export included. |
| Languages and norms | 6 languages with informal register: Spanish, English, French, German, Italian, Portuguese (Brazil). Empirical percentiles normed by country where data exists. | Broad multilingual support. No country-level psychometric norms: there is nothing to norm. |
| Company behind | Bilbao AI S.L. — Spanish company, Holded invoicing with TicketBAI/BATUZ. Privately held, no VC rounds. An identifiable human to talk to. | OpenAI, US company. Unusual corporate structure (capped-profit), Microsoft investment, US jurisdiction. Human contact hard to reach for end users. |
Where ChatGPT wins
No sugarcoating. Selling a cognitive profile requires admitting other products do some things better. Here they are.
Breadth of use cases
ChatGPT covers scenarios Afini does not aim at: image generation, code interpreter, canvas, massive file analysis, voice. Afini is specific — AI personalized with your profile. For everything else, ChatGPT wins on breadth.
App and plugin ecosystem
ChatGPT plugs into Zapier, n8n, Slack, Teams, Notion, VS Code, Office. Afini connects via API and portable JSON — enough for developers, less plug-and-play for non-technical users.
Where Afini wins
Verifiable claims. No "we are the only ones who truly understand you". Numbers, formats, protocols.
Quantified profile, not fragments
ChatGPT remembers "you like football, you have a cat, you prefer Python over JavaScript". Afini knows you are O=67, N=40, securely attached, ironic humor. The difference is the one between a post-it and a CT scan.
You take your profile with you
Afini’s portable JSON (€249 one-time or included in the annual plan) can be injected as a system prompt into any LLM. ChatGPT offers no equivalent; the day you leave, the memory stays.
Anti-sycophancy contracted
Afini injects 5 explicit anti-flattery directives into every conversation’s system prompt, calibrated by your profile. ChatGPT suffers from the phenomenon (admitted by OpenAI in April 2025) and handles it via model updates, not user control.
What each plan costs in Afini
Questions you are asking yourself
- Can I import my ChatGPT conversations?
- Not directly, and it would not make sense either: ChatGPT memories are model-selected fragments, not a structured profile. Afini builds your profile from scratch with tests and guided conversations, measuring things ChatGPT never asked.
- Does Afini replace ChatGPT in my daily life?
- Depends on the use. For personal decisions, meaningful conversations and anything where "the AI knowing you" adds value, yes. For generating a logo image or running code, no — Afini focuses on the cognitive profile. Many users keep both.
- What happens to my profile if I cancel?
- You can export it before canceling. The portable JSON (€249 one-time or bundled with the annual plan) works in any LLM for years. If you cancel without exporting, your profile is kept 30 days to be recovered; afterwards it is deleted.
- Do my Afini conversations train any model?
- No. Not Afini, not Anthropic. It is part of the user contract and of the contract Afini has with Anthropic (Zero Data Retention on model calls).
- Why does ChatGPT flatter me?
- Because it was trained to please you, not to be useful. RLHF rewards responses human raters mark as satisfying, and those tend to be flattering. Afini mitigates this by injecting an anti-sycophancy block calibrated by your profile; if your A and N are high, the anti-flattery kicks in harder.
Your profile in portable format. In any AI.
€14.99 a month. Cancel whenever. If it does not convince you, say so.
No signup up front. First you decide, then you pay, then you identify yourself.