Paranormal investigators are adding AI tools to kits that already relied on phone sensors and word-generating apps. The Guardian reports that some enthusiasts worry this makes it harder to tell what counts as evidence.
Key takeaways
- The Guardian reports that ghost hunting has changed in recent years because of AI chatbots, and that some paranormal enthusiasts are uneasy about it.
- One app, GhostTube Seer, is promoted by its makers as the world’s first AI ghost-hunting tool.
- Paranormal investigators have long taken scientific instruments such as EMF meters and used them to look for signs of the supernatural.
- Language models can produce fluent, plausible text from almost any input, so an AI reply can feel meaningful whether or not anything unusual is happening.
- AI ghost hunting raises questions about evidence, interpretation and trust that also come up in mainstream debates about generative AI.
What is happening in ghost hunting?
Amateur paranormal investigation now relies heavily on technology. According to the Guardian, two developers who investigate places said to be haunted bring software they built themselves. One of their apps takes readings from a phone’s sensors and matches the signals to words stored in a database. In practice, the phone produces words that users may read as messages.
The pair have now started working with AI chatbots. The Guardian reports that they describe their app GhostTube Seer as the world’s first AI ghost-hunting tool. The source material does not explain in detail how the app works, which model it uses or what it shows users, and those details are not known here.
This is not a sudden break with the past. Ghost hunters have used electronic devices for decades, and the newer tools build on that habit. What has changed is the kind of output. Older devices produced beeps, static or single words. A system built on a large language model can produce full sentences that sound like conversation.
Why is AI ghost hunting in the news now?
The topic is in the news because the Guardian has published a feature on how the spread of AI chatbots is affecting the paranormal community. It appeared around Halloween, when interest in ghosts and haunted places usually rises.
There is also a broader reason. Since ChatGPT and similar systems became widely available, generative AI has spread into nearly every hobby and profession. Ghost hunting is an unusual case because it depends so heavily on interpretation: whether an output is meaningful is often the very thing under dispute. The Guardian’s reporting suggests that many enthusiasts have doubts about AI, not just the people building these tools. The source material does not say exactly what worries them or how many share those concerns.
What background does a newcomer need?
Ghost hunting is the informal practice of visiting places said to be haunted and trying to record evidence of spirits. It became popular through television and online video, and today much of it happens through social media and phone apps.
Its basic toolkit borrows from science. An EMF meter detects electromagnetic fields. Engineers and electricians use these meters for practical work, such as finding wiring faults. Paranormal investigators use them in a different way, treating sudden changes in a reading as a possible sign that a presence is nearby. The Guardian describes this as scientific tools being repurposed.
Other common methods include:
- Electronic voice phenomena (EVP): listening to audio recordings for voices that were not heard when the recording was made.
- Spirit boxes: devices that sweep quickly through radio frequencies and produce bursts of sound that listeners may hear as words.
- Word-generator apps: software that converts sensor readings or random values into words chosen from a fixed list.
Mainstream science does not accept any of these methods as evidence of spirits. Psychologists point to well-documented tendencies such as pareidolia, which is seeing or hearing meaningful patterns in random input, and to the way expectations shape what people perceive. Faint, noisy or brief signals are especially easy to interpret in a way that fits what the listener hopes to find.
Generative AI adds a new element. Large language models work by predicting likely text. They are designed to produce coherent, relevant-sounding answers, and they can confidently generate content that has no factual basis, a problem often called “hallucination”. When a model’s input is tied to sensor noise, the result can read like a clear reply even though it reflects how the model works rather than anything outside the phone.
Who is affected, and how?
The most direct effect is on hobbyists. People who use AI-powered apps on investigations may get responses that feel far more convincing than a crackle from a spirit box. That can make an investigation more exciting. It can also make it harder for people to tell apart a reading they produced themselves, one the software produced, and something they cannot explain.
Developers of paranormal apps are affected too. Adding AI features is a way to stand out in a crowded app market, but it puts them in a debate about transparency: what the app actually does with sensor data, and how clearly that is explained to users.
Audiences for paranormal videos and podcasts are another group. If content creators use AI tools on camera, viewers may struggle to judge what they are seeing. Generative AI can also produce convincing fake images, audio and video, which may make claimed “evidence” shared online even harder to assess.
Finally, the topic matters to people who have no interest in ghosts. How people react to AI outputs in a haunted building is a small-scale example of a wider question: how readily people attribute intention and meaning to machine-generated language.
Where do informed people disagree?
Disagreement runs on several levels. Within the paranormal community, some people see AI as one more tool in a long line of devices, useful for exploration and entertainment. Others, as the Guardian’s reporting indicates, are worried. One likely reason is that a tool which produces fluent language on demand could flood investigations with output that looks like communication. The specific arguments enthusiasts make are not fully set out in the source material.
Sceptics and scientists tend to take a more basic position. Because there is no accepted evidence that existing ghost-hunting devices detect spirits, they argue, an AI layer adds more convincing noise rather than better detection. From this view, the main risk is that AI exploits normal human pattern-seeking more effectively than older tools.
A third position treats ghost hunting mainly as a form of entertainment or storytelling. People who hold it may see AI apps as harmless, much like a game or a themed attraction, as long as users understand what the software is doing. The disagreement then shifts to whether apps describe themselves honestly.
What are the practical implications?
For anyone using these tools, it helps to know some general facts about AI systems. A language model will usually produce an answer when prompted, whatever the input. Its outputs can be shaped by the words or questions a user supplies. And fluency is not evidence of accuracy.
Investigators who want their work to hold up to scrutiny may find it useful to record exactly which app and settings they used, to note when outputs came from software rather than direct observation, and to be cautious about presenting AI-generated text as a message from anyone.
For app makers, the questions concern disclosure: explaining whether outputs come from a general-purpose AI model, how sensor data feeds into it, and what the limits are. Whether particular apps already do this is not known from the source material.
For the wider public, AI ghost hunting is a reminder to treat striking AI-generated content with care, especially when it is shared online without context.
What should people watch next?
Several developments are worth following. One is whether more paranormal apps add generative AI features, and whether app stores or consumer-protection rules have any bearing on how such apps describe their capabilities. Another is how paranormal communities respond. Some may set their own norms for AI use in investigations, while others may adopt the tools fully.
More broadly, research into how people interpret and trust AI-generated language is likely to stay relevant. Ghost hunting is a niche setting, but it shows clearly how people respond when software produces confident answers to questions that have no verifiable answer.
Frequently asked questions
What is GhostTube Seer?
GhostTube Seer is an app that its makers describe as the world’s first AI ghost-hunting tool, according to reporting by the Guardian. It comes from developers who already build paranormal investigation software, including an app that matches phone sensor signals to words in a database. Public details of exactly how Seer uses AI, and which underlying model it relies on, are not covered in the material available for this article.
Can AI actually detect ghosts?
There is no scientific evidence that AI, or any other technology, can detect ghosts. Mainstream science does not accept spirits as a measurable phenomenon. AI language models generate text by predicting likely words, so they can produce responses that sound meaningful from almost any input, including random sensor readings. A convincing reply from an AI tool shows how the software works, not that it has contacted anything.
What is an EMF meter used for in ghost hunting?
An EMF meter measures electromagnetic fields. Electricians and engineers use it for practical tasks such as checking wiring. Paranormal investigators use it differently, treating unexpected changes in readings as possible signs of a presence. Sceptics point out that ordinary sources such as electrical wiring, appliances and phones commonly cause these changes, which makes such readings unreliable as evidence of anything supernatural.
Why are some ghost hunters worried about AI?
The Guardian reports that some paranormal enthusiasts are concerned about how AI chatbots are changing their hobby, though the source does not set out every reason. A common worry about generative AI is that it produces fluent, confident output whatever the input. In ghost hunting, that could make it harder to tell genuine puzzles apart from text the software generated, and make online evidence harder to judge.
What is pareidolia and why does it matter here?
Pareidolia is the human tendency to see or hear meaningful patterns, such as faces or words, in random or vague input. It helps explain why people may hear voices in static or find messages in random word lists. AI tools that turn noise into fluent sentences could strengthen this effect, because the output already looks like structured communication before anyone interprets it.
Is using AI apps for ghost hunting harmful?
For many people, AI ghost-hunting apps are a form of entertainment, and using them carries little direct risk. Concerns come up when users do not understand that the software generates its own text, or when AI output is presented online as proof of the supernatural. Knowing how language models work, and treating their responses with caution, helps keep the activity in perspective.
Sources and further reading
- The Guardian technology section: reporting on how AI chatbots are changing ghost hunting and how paranormal enthusiasts are responding.
- Academic psychology literature on pareidolia, perception and belief in the paranormal.
- Technical documentation from AI developers explaining how large language models generate text and why they can produce inaccurate output.
- Guidance from electrical safety and engineering bodies on what EMF meters measure and common sources of electromagnetic fields.
Surfaced from the rss:guardian_tech signal “AI in ghost hunting”. AI-assisted draft, editorially reviewed.
