Online dating depends on people being able to build trust through profiles and conversation. Avast researchers warn that LoveGPT, a tool designed to assist with dating interactions, can also be used to automate fake personas and draw users into scams.
More convincing conversations, at greater scale
According to Avast, the software behind LoveGPT has existed for more than a decade. OpenAI language models were added in March 2023, making its profiles and messages more believable, the researchers say.
Scammers can use fake profiles and flirtatious messages to persuade people to share sensitive information, including credit card details. Such schemes predate today’s language models, but automated tools can help operators reach and engage more people.
LoveGPT uses GPT-3 and GPT-3.5 to write profile descriptions and exchange messages with users. Avast says it can tailor its presentation to the conventions of different dating services—for example, posing as a poet on Bumble or expressing enthusiasm for travel on Zoosk.
Automation and concealment features
The concern extends beyond the text a user sees. Avast reports that LoveGPT can create many accounts, send and receive email, and interact with people on dating platforms. It can also bypass captchas and mask access using proxies and browser anonymization tools.
Together, these functions can make an automated account appear more like an ordinary participant. Natural-sounding messages may be harder to recognize as machine-generated, while platform-specific prompts can help a fake profile fit the setting.
Avast lists the following platforms as compatible with LoveGPT:
- Ashley Madison
- Badoo
- Bumble
- Craigslist
- DuyenSo
- Facebook Dating
- likeyou.vn
- MeetMe
- OkCupid
- Plenty of Fish (POF)
- Tagged
- Tinder
- Zoosk
What users can do
Avast advises people to be cautious with new online contacts and avoid sharing personal information, even when a profile or conversation seems authentic. That guidance matters because a fluent, engaging exchange is not proof that the person behind an account is who they claim to be.
Users can treat requests for sensitive details with care and take time to assess a new contact. The source does not describe a single telltale sign that reliably identifies LoveGPT, so caution should not depend on spotting awkward or obviously automated messages.
Questions about access remain
The report raises a question about how a tool associated with dating scams can access OpenAI language models. The Decoder said it had asked OpenAI whether LoveGPT’s use of its models violates OpenAI’s content policy, and that it would update its article if it received a response.
Avast’s warning points to a broader challenge for online dating: language models can make deceptive interactions more convincing, while automation can increase their reach. For users, the practical advice is straightforward—stay guarded with personal information, even when an online conversation feels natural.