You capture a photo during a session, open it in the gallery, and tap "Analyze." A few seconds later, GhostTech hands you back a verdict, a list of signals, and a number from 0 to 100. What actually happens in those seconds? In this article we open the box: what gets sent, where it goes, what each part of the result means and — most importantly — what it doesn't mean.

First things first: what it is and isn't

  • AI analysis is an optional feature: it never happens automatically or in the background, only when you pick a photo in the gallery and tap "Analyze."
  • It's the only feature in GhostTech that needs internet: EMF radar, SLS camera, EVP, night vision, geophone, and surveillance mode all process everything on your device, fully offline.
  • The verdict is an interpretation from a language model, not a scientific measurement or proof of paranormal activity.

What actually gets sent to the AI

When you capture a photo during an investigation, GhostTech automatically links that image to a "snapshot" of your sensors at that exact instant: the EMF radar reading, the geophone level (if the phone was still), the audio state, and camera info. That photo and that snapshot are saved together in your gallery, on your device, with nothing leaving your phone yet.

Only when you open the gallery, pick that specific photo, and tap "Analyze" does the send happen: the image and its linked sensor snapshot travel to Gemini 2.5 Flash, Google's AI model, through Firebase AI Logic. Nothing else from your session gets sent — not your other photos, not your surveillance history, not your location — and nothing is ever sent without you explicitly asking for it.

CaptureYou take the photo; GhostTech automatically links the sensor snapshot from that exact instant, all local.
AnalyzeYou tap "Analyze" in the gallery: only then does the photo + snapshot travel to Gemini 2.5 Flash (Google).
VerdictThe AI returns an interpretive text, a list of signals, and a 0-100 level.

What a real verdict looks like

The result the AI hands back is internally called an AiReport, and it has three parts. Here's a simplified version of how it's presented:

AiReport — Illustrative example

Gemini 2.5 Flash

"The image shows a partial silhouette next to the window, coinciding with an EMF spike recorded in the same second. The vibration reading isn't reliable in this capture because the phone wasn't completely still."

0
EMF spike of 71 µT, simultaneous with the capture.
Partial outline detected in the upper-left area.
Geophone reading discarded: the phone wasn't stationary.

The verdict is the natural-language summary: what the AI "sees" in the image and how it relates that to the sensor readings from the same instant. The signals are the list of specific clues backing up that verdict — including the ones it discards, as in the example, where the AI itself explains that the geophone data isn't valid because the phone was moving. The 0-100 level sums up the combined intensity of every signal found: it's not a probability of a ghost, it's a score of how many interesting coincidences were detected.

A language model, not a sensor

Gemini interprets image + data, but it doesn't "detect ghosts": it generates the most plausible reading of visual and numeric patterns, the same way a human investigator would looking at the same photo.

On-demand, never in the background

It only activates when you tap "Analyze" on an already-saved photo. The rest of the app keeps running 100% offline, on your device.

Free, no API key in the app

GhostTech uses the free tier of Firebase AI Logic for Gemini 2.5 Flash. If the service gets overloaded, the app automatically retries up to 3 times before warning you of an error.

Why it sometimes takes a while or fails

Gemini 2.5 Flash through Firebase AI Logic runs on a free tier shared by every user of the app. During high-traffic moments it can get overloaded for a few seconds. If that happens, GhostTech automatically retries up to 3 times, with a short wait between attempts, before showing you an error. In the vast majority of cases the verdict arrives within a few seconds.

What a high level means (and doesn't mean)

A level of 80 or 90 isn't a certificate of paranormal activity: it means the AI found several matching signals in that particular photo — a partial silhouette, a simultaneous sensor spike, an unusual pattern — and summarized them into a number. The cause behind those signals can have a perfectly mundane explanation: a reflection, a shadow, electromagnetic interference from a nearby appliance, a vibration from the building. Treat the verdict as a second pair of eyes that helps you prioritize which photos deserve a closer look, not as a final result.

To dig deeper into interpreting each type of signal on its own, we have dedicated guides on how to read EMF meter readings and on what orbs in photos really are.

0parts of a verdict
0automatic retries if overloaded
0photos analyzed without your permission

Your privacy during analysis

AI photo analysis is the only exit door for personal data anywhere in the app: while the EMF radar, the Ghost Vision SLS camera, the geophone, the EVP engine, and surveillance mode process absolutely everything on your device, this specific feature needs to send a photo to a Google server so it can be interpreted. That's why it's designed to be explicit and one-off: one button, one photo, one send, one response. You can review every detail — including the anonymous usage analytics we also use to improve the app — in our privacy policy.

Try AI analysis on your next session

EMF radar, Ghost Vision SLS camera, EVP engine, night vision, digital geophone, and now AI photo analysis: all in one free Android app.

What would you like the AI to do next?

Would you like the AI to compare several photos from the same session, or explain each signal in more detail? Let us know.

FAQ about AI analysis

What is GhostTech's AI photo analysis?

It's an optional feature inside the gallery: you pick a photo you've already captured, tap "Analyze," and GhostTech sends that image together with the sensor readings taken at the exact same moment to Gemini 2.5 Flash, Google's AI model, to generate a verdict with detected signals and a 0-100 anomaly level.

What data is actually sent to the AI?

Only the photo you choose to analyze and the sensor snapshot (EMF, geophone, audio, camera state) that was linked to that photo at the moment of capture. No other data from your session, and never in the background: the analysis only happens when you tap "Analyze."

What do the verdict, the signals, and the 0-100 level mean?

The verdict is a natural-language summary of what the AI observes in the photo and its linked readings. The signals are the specific clues it found. The 0-100 level sums up the combined intensity of those signals, not a scientific probability of a paranormal event.

Does the AI's verdict prove there's a ghost in the photo?

No. It's an interpretation of visual and sensor patterns generated by a language model, not scientific proof. A high level only means the AI found several matching signals; the explanation for those signals can be perfectly mundane.

Do I need internet to analyze a photo with AI?

Yes, but only for that one-off step: GhostTech needs a connection to send the photo and readings to Gemini and get the verdict back. The rest of the app — EMF radar, SLS camera, EVP, night vision, geophone — doesn't need internet because it processes everything on your device.

Conclusion

AI photo analysis doesn't replace your own judgment: it gives you a second, structured point of view, linked to the sensors from the exact instant of capture, so you can decide better which photos deserve a closer look. It's one-off, it's optional, and it's the only part of GhostTech that ever leaves your phone — and only when you ask for it.

Remember: the AI's verdict is an entertainment and interpretation-support tool, not a scientific measurement. No score, however high, constitutes proof of paranormal activity.