Hayati AI Review

What Hayati AI is

Hayati AI is an Arabic-first AI girlfriend app run by Hayati Labs Limited, registered in Hong Kong. Its characters speak regional Arabic, including Najdi, Hijazi, Gulf, Iraqi and Egyptian dialects, rather than formal Modern Standard Arabic. Chat starts free, and voice notes, photos and short private clips are paid with Lulu credits or a monthly subscription. Prices are not published before signup. The privacy policy allows precise device-location logging where permission already exists and infers your dialect from chats, while the safety policies ban minors, non-consensual content and real-person impersonation.

What it costs

Pricing is published as Free to start (paid prices hidden until signup).

On credits: Lulu (pearls) credits pay for photos, voice and extra features. Non-refundable, but unused credits never expire.

Privacy and data

Privacy scores 4.5 out of 10 here, read off the published policy rather than the marketing page.

• The account can be deleted from settings.
• Chat history can be deleted.
• Signup is not anonymous.
• No phone number is required.
• Conversations are stored on their servers.
• Your data may be used to train their models.

Verified against the Hayati AI privacy policy effective 30 August 2026. Against it: where device location permission was already granted it may record precise coordinates without a new prompt, and it keeps an approximate login-location history from your IP. It infers your likely Arabic dialect from your chats and compares product use and payment attempts across dialect groups. User data is used to develop its AI models and deleted data may persist in anonymised form for training (PII is said to be excluded). Data is processed in the United States and Hong Kong. In its favour: rights to access, correct, delete and export data, object to profiling and opt out of marketing; GCC PDPL, EEA, UK and California rights acknowledged; no language about selling personal data; phone not required at signup.

Source: the published privacy policy.

Content and safety policy

• Characters are stated to be fictional adults only.
• There is an age gate on entry.
• Real-person likenesses: Fictional characters only. The content monitoring policy says there are no real profiles, human performers or user-to-user interactions, and bans impersonation or depiction of real, identifiable individuals without authorization, alongside any content involving minors and non-consensual content.

Source: their published terms.

How the filter behaves in a live conversation is a separate question, and one only a hands-on session can answer.

Who should skip it

Anyone who wants to see prices before signing up, needs a native iOS or Android app, or is uncomfortable with precise location logging and dialect profiling.