Development and Application of a Specialized AI Agent to Analyze UAP Reports

 

Murray Toews, Ufology Research

Chris Rutkowski, Ufology Research

Winnipeg, Manitoba, Canada

 

The relatively recent development of AI Agents to assist researchers in detailed studies within their fields has seen many applications in business, engineering, and science. The creation of Agents has been seen as a godsend resource that could revolutionize analyses across many disciplines, including ufology.

The development of AI Agents represents one of the most monumental shifts in technology because it marks the transition from AI as a passive assistant to AI as an active collaborator. AI Agents do not need step-by-step instructions; they simply require a high-level objective, then break it down into a series of smaller tasks.

Beyond daily productivity, specialized AI Agents are accelerating breakthroughs in complex fields. They are being deployed to run virtual chemistry labs, automate the testing of new materials, scan billions of data points for medical research, and model climate change scenarios—essentially acting as tireless, 24/7 research partners for human scientists.

With this in mind, an AI Agent has been developed specifically for the analysis of UAP reports submitted for consideration. During the past 75 years, ufologists have created and applied several reporting forms for inputting case data, then used the data in analyses of the UFO phenomenon. This has often been arduous and time consuming, requiring the consideration of a multitude of factors and possible explanatory pieces of information. It has seemed reasonable that an AI Agent could automate and accelerate the investigative and analyze UAP process.

As an early experiment in using AI in ufology, the Canadian UAP Monitoring Agent Framework (CUMAF) was developed by Ufology Research to test how AI could be used within this field of study. It was designed as a practical workflow for collecting, organizing, and reviewing Canadian public UAP/UFO postings across social media platforms.

Purpose

This framework is designed to support a repeatable, evidence-conscious review of publicly available Canadian UAP/UFO-related postings. It is intended for independent research, potential Canadian UFO Survey support, and future public-facing projects such as talks, screenings, and research summaries.

The framework does not attempt to prove or disprove extraordinary claims. Its purpose is to capture postings, preserve links, standardize basic metadata, and prioritize which items deserve deeper review.

Primary Scope

·        Country: Canada only

·        Initial platform focus: YouTube and publicly indexed video posts

·        Time period: monthly collection, with optional backfill searches

·        Case status: preliminary until reviewed against astronomical, aviation, meteorological, satellite, and local context

Recommended Source Tiers

Tier 1: YouTube/public video

Accessible, linkable, and easier to search. Useful as the first monitoring layer.

Tier 2: Public news and institutional reports

Useful for corroboration, context, and media coverage.

Tier 3: Public social media

Reddit, Bluesky, X/Twitter, public Facebook posts or groups, where accessible and ethically usable.

Tier 4: Formal reporting databases

Canadian UFO Survey, NUFORC, MUFON or other databases where permitted and appropriate.

Core Data Fields

Credibility Score Guide

Strangeness Score Guide

Monthly Workflow

1.        Run Canada-wide searches at the start of each month for the previous month.

2.        Search by country, province, city, and terms such as UFO, UAP, strange lights, orb, triangle, fireball, OVNI, and unexplained aerial.

3.        Log everything first, then filter for relevance.

4.        Separate witness-posted material from commentary, news coverage, podcasts, and historical/archival content.

5.        Flag likely explainable cases, but do not discard them without noting the likely explanation.

6.        Prioritize cases with date/time, location, original media, multiple witnesses, or flight/astronomy check potential.

Suggested Monthly Search Strings

·        "UFO Canada 2026"

·        "UAP Canada 2026"

·        "UFO sighting Canada [month] 2026"

·        "strange lights Canada [month] 2026"

·        "UFO Ontario 2026" / "UFO Manitoba 2026" / province variations

·        "OVNI Quebec 2026"

·        "fireball Canada [month] 2026"

·        "strange object by moon Canada 2026"

Review Notes

Each item should be treated as a lead, not as a confirmed UAP case. Before a case is promoted to Reviewed or Investigated, check astronomy apps, aircraft data, satellite visibility, meteor/fireball reports, weather, camera artifacts, repost status, and local media corroboration where possible.

Useful Context Sources

·        Sky Canada Project - Office of the Chief Science Advisor: https://ised-isde.canada.ca/site/science/en/office-chief-science-advisor/sky-canada-project

·        Sky Canada final report: https://science.gc.ca/site/science/en/office-chief-science-advisor/sky-canada-project/management-public-reporting-unidentified-aerial-phenomena-canada

·        Canadian UFO Survey: https://canadianuforeport.ca/

·        Library and Archives Canada UFO records: https://www.canada.ca/en/library-archives/collection/research-help/science-technology/ufos.html

 

It was noted that adding or subtracting parameters such as core data fields, search strings, and sources would yield differing results. The versatility of such an approach suggested that the use of the AI Agent would be beneficial to ufologists seeking information on UAP cases not reported through regular UFO reporting systems or organizations.

In the first application of CUMAF, a test run identified a number of UAP cases for consideration. The AI Agent looked only at YouTube and other public web sources between January 1, 2026, and June 4, 2026.

 

Scope and Important Limits

This is a first-pass backfill, not a complete census of all Canadian UFO/UAP postings. YouTube search results are incomplete, change over time, and may not expose every upload, Short, deleted video, private video, or low-indexed post. Several entries below are leads based on publicly indexed titles/snippets rather than full video review.

The scoring is deliberately conservative. A low score does not mean the witness is wrong; it means the available public metadata is thin or the item may have a conventional explanation. A high score should be treated only as a priority for follow-up.

Summary of Results Found

Preliminary Case Table

Priority Follow-Up Leads

·        Sudbury, Ontario black triangle video - because it includes a date, time, and location in the title.

·        Winnipeg-to-Calgary flight object - because an incident report is referenced and may be traceable.

·        B.C. strange shape news clip - because multiple residents/media coverage may make corroboration easier.

·        Pembroke Mother's Day sighting - because the date can likely be narrowed and checked against sky/flight data.

Suggested Next Checks for Each Case

7.        Confirm exact upload date and whether the video is original or reposted.

8.        Capture video title, channel name, description, comments that contain witness details, and any stated date/time/location.

9.        Check sky conditions, visible planets, Moon position, meteor/fireball reports, satellite passes, aircraft tracks, and local event/weather data.

10.    Separate likely identified cases from unresolved cases rather than deleting them; identified cases are useful for education and baseline comparison.

Source Index

·        Sudbury triangle: https://www.youtube.com/watch?v=414yCcxx0LA

·        Winnipeg-Calgary flight object: https://www.youtube.com/watch?v=hgZ_2Y0_6d4

·        Vancouver Feb. 19 object: https://www.youtube.com/watch?v=sk4IcaBoT2g

·        Manitoba UFO researcher / AI tracking: https://www.youtube.com/watch?v=Ej6mgDv2Yms

·        Increase in unexplained aerial sightings across Manitoba: https://www.youtube.com/watch?v=fsyMddjvIs0

·        Vancouver Island celestial object: https://www.youtube.com/watch?v=IQLupGIJMSU

·        NWT drillers cloud claim: https://www.youtube.com/watch?v=d9hlH-L8Jdo

·        Pembroke Ontario Mother's Day: https://www.youtube.com/watch?v=R91ai0DNTxs

·        Ontario apartment UAP: https://www.youtube.com/watch?v=-ziP1-p5f1E

·        Manitoba May 8 Short: https://www.youtube.com/shorts/sJKy0TUDz8A

·        BC orb/cloud claim: https://www.youtube.com/watch?v=bs8IiMFvLdI

·        BC strange shape news clip: https://www.youtube.com/watch?v=4_a_WjHCaHY

 

The ease with which the AI Agent searched for and located UAP case sources was deemed a success, and bode well for adapting the technology for other uses within ufology.

The next logical step was an application of an AI Agent in the analyses of UAP reports themselves. Often, a researcher or investigator will receive a photograph or video from a UAP witness and be requested for an opinion or (more usually) an explanation. The process or procedure for these investigations has been laid out and documented many times, especially in manuals prepared by UFO organizations for use by its members and field investigators. The automation of this process would be of major benefit in the study of reports and in case investigations.

To this end, an AI Agent was created to employ standard photographic and image analysis techniques for UFO/UAP investigations.

Its purpose was to outline commonly used photographic, video, and digital forensic techniques that can assist investigators in evaluating photographic and video evidence associated with UFO/UAP reports. These methods focus on objective analysis rather than interpretation and might help identify conventional explanations, detect image manipulation, or strengthen the evidentiary value of a case.

The steps considered by the AI Agent were:

1. Metadata Analysis (EXIF)

 Objective: Extract technical information embedded within image and video files.

 Analyses of available case information:

• Date and time captured
• Camera or smartphone model
• Lens information
• Exposure settings
• GPS coordinates (when available)
• Editing software history

Application:

• Verify witness claims
• Confirm location and timing
• Detect post-processing or editing

2. Error Level Analysis (ELA)

Objective: Identify inconsistencies in image compression.

Application:

• Detect pasted or inserted objects
• Identify composite images
• Highlight areas of possible manipulation

Limitations: Results become less reliable after repeated social media compression.

3. Noise Pattern Analysis

Objective: Compare digital sensor noise throughout an image.

Application:

• Determine whether an object shares the same image characteristics as the surrounding scene
• Detect possible image insertions

4. Sensor Pattern Noise (PRNU)

Objective: Analyze the unique fingerprint generated by a camera sensor.

Application:

• Verify image origin
• Detect compositing
• Support forensic authentication

5. Pixel-Level Edge Analysis

Objective: Examine object boundaries and transitions.

Application:

• Detect artificial edges
• Identify cut-and-paste artifacts
• Evaluate image consistency

 6. Motion Blur Analysis

 Objective: Evaluate blur patterns within photographs or video frames.

 Application:

• Determine object motion
• Assess camera movement
• Differentiate between moving objects and camera shake

 7. Bokeh Analysis

 Objective: Examine out-of-focus light sources.

 Application:

• Identify stars, planets, aircraft lights, and distant light sources

NB: This could explain many common “orb” sightings

8. Focus-Distance Analysis

Objective: Estimate whether an object is near or distant relative to the camera.

Application:

• Identify possible insects in image(s)
• Identify possible airborne debris
• Distinguish foreground objects from distant targets

9. Lens Flare Analysis

Objective: Detect optical reflections occurring inside a camera lens.

Application:

• Identify lens "ghosts"
• Explain apparent luminous objects
• Evaluate sunlight-related artifacts

10. Reflection Analysis

Objective: Analyze reflective surfaces and reflected light paths.

Application:

• Detect vehicle windshield reflections
• Identify window reflections
• Evaluate water reflections

11. Shadow and Lighting Analysis

Objective: Assess consistency of illumination within a scene.

Application:

• Determine whether an object is lit realistically
• Identify possible CGI or compositing

12. Atmospheric Perspective Analysis

Objective: Evaluate how atmosphere affects distant objects in the image(s).

Application:

• Estimate relative distance to object(s)
• Assess haze, color shift, and contrast reduction
• Determine whether an object's appearance is physically plausible

13. Photogrammetry

Objective: Use geometric measurements to estimate object characteristics.

Application:

• Estimate size
• Estimate altitude
• Estimate distance
• Estimate speed

NB: This is one of the most powerful tools available when suitable reference points exist.

14. Angular Measurement Analysis

Objective: Calculate an object's apparent position using known landmarks.

Application:

• Determine azimuth and elevation
• Estimate object location
• Support trajectory reconstruction

15. Frame-by-Frame Video Analysis

Objective: Analyze motion and behavior across video frames.

Application:

• Measure speed of object(s)
• Track direction changes
• Detect rotation and acceleration

 16. Video Stabilization Analysis

 Objective: Remove camera shake from video footage.

 Application:

• Reveal true object motion
• Eliminate apparent erratic movement caused by hand-held recording

17. Optical Flow Analysis

Objective: Track pixel movement between video frames.

Application:

• Determine motion vectors
• Estimate trajectories
• Compare object motion to background motion

18. Spectral and Color Analysis

Objective: Examine color channels and spectral characteristics.

Application:

• Evaluate lighting sources
• Identify LED signatures
• Assess infrared contamination
• Analyze atmospheric effects

19. AI-Generated Image Detection

Objective: Identify evidence of synthetic image generation.

Application:

• Detect AI-generated hoaxes
• Identify diffusion-model artifacts
• Detect unrealistic geometry or texture generation

20. Reverse Image Search

Objective: Determine whether an image has appeared online previously.

Application:

• Detect recycled imagery
• Identify misattributed photographs
• Locate original sources

21. Astronomical Correlation

Objective: Compare observations against known celestial objects and events.

Sources:

• Planets
• Bright stars
• ISS
• Starlink
• Meteor activity
• Known satellite passes

Application:

• Eliminate common astronomical misidentifications

22. Aircraft Correlation

Objective: Compare reports with known aircraft activity.

Application:

• Verify commercial traffic in area
• Identify helicopters in area
• Compare witness observations against available flight records

23. Weather and Atmospheric Correlation

Objective: Assess environmental conditions during the observation.

Factor:

• Cloud cover
• Temperature inversions
• Lightning
• Aurora activity
• Lenticular cloud formations

Application:

• Identify atmospheric explanations

24. Geospatial Reconstruction

Objective: Reconstruct a sighting using terrain and location data.

Application:

• Determine lines of sight
• Estimate object position
• Recreate witness perspective

 

Given these objectives, the AI Agent was given a suggested core workflow for UFO/UAP analyses:

1. Metadata Extraction

2. Reverse Image Search

3. Error Level Analysis

4. Lens Flare Detection

5. Bokeh Analysis

6. Motion and Stabilization Analysis

7. Photogrammetry

8. Astronomical Correlation

9. Aircraft Correlation

10. Weather Correlation

11. AI-Image Detection

12. Confidence Scoring

 

The expected output structure would therefore be, as an example:

 

Category| Result

Metadata| Consistent

Evidence of Editing| None Detected

Astronomical Match| No Match

Aircraft Match| No Match

Weather Explanation| Unlikely

AI Detection| Low Probability

Confidence Level| Moderate

Classification| Unresolved

 

The guiding principle in the use of the AI Agent in such an analysis is that a properly-designed UFO/UAP analysis system should not attempt to prove or disprove extraordinary claims. Instead, it should objectively evaluate available evidence, identify conventional explanations where supported by data, and document unresolved cases using transparent and repeatable methods.

 

Application of the AI Agent

The next step in the development of the AI Agent was to test it on a UAP case in order to evaluate its effectiveness. An opportunity recently arose when a witness to a UAP in Italy emailed details of his sighting to many UFO investigators and researchers around the world. This seemed to be an ideal case to have the AI Agent examine. Further, the witness invited recipients to investigate the case, so this seemed a good candidate, particularly since he provided an extraordinary amount of detail of the observation as well as significant video and photographic evidence that would be necessary for a proper evaluation. (Ironically, the witness admitted the use of AI in the preparation of his report.)

The case, as received from the witness on June 9, 2026, is as follows:

My name is [name deleted] and I would like to report the sighting of an unidentified flying object observed in the area of Acquaviva Picena, in the province of Ascoli Piceno, near the Medieval Fortress of Acquaviva Picena.

During the observation, I saw a bright, intense white luminous object, apparently silent, moving slowly and following an unusual trajectory. The object appeared to move along the area between the historic center / the fortress and the fields, with a possible change in altitude or final disappearance.

I have prepared documentary material related to the sighting, including:

·        video of the observation;

·        annotated video extract;

·        PDF/DOCX report;

·        approximate coordinates of the observed trajectory;

·        images and video frames;

·        preliminary technical analysis of the video. 

I would like to clarify that I am not claiming any specific origin of the object. I kindly request a verification to understand whether it could have been air traffic, a drone, a helicopter, a satellite, an atmospheric phenomenon, or another known object.

I would also like to clarify that I am sending this report because the event made a strong impression on me: the object appeared directly in front of my windows and was clearly visible.

I do not intend to state with certainty the nature or origin of the object. However, based on what I observed, it did not seem similar to a normal drone, an aircraft, or a common phenomenon. It appeared as a very bright sphere of intense white light, almost similar to a white firework, but with a movement that seemed controlled and not random.

For this reason, I kindly request a technical verification by the competent authorities, in order to understand whether the object can be explained as air traffic, a drone, a satellite, an atmospheric phenomenon, or another known event.

I would also like to specify that this message, the report, and the organization of the information were prepared with the support of ChatGPT, based on my personal observation, the recorded video, and the materials collected.

I used this tool only to better organize the data, translate the text, and present the information more clearly. The observation described is real and concerns what I personally saw in front of my windows.

I hope that the material sent may be useful for a technical verification or for any further investigation by your organization.

https://drive.google.com/drive/u/0/folders/1cOIHMdw8uzQ8Dzma1rr03nFICSaz80Q6

Witness Observation Story:

The witness reports that he was sitting at his computer in the evening, playing games. His windows face the valley near Acquaviva Picena. In the background, he saw what initially appeared to be a firework or flare launched in the valley, visually close enough and bright enough that he expected an explosion. Instead, the apparent firework-like object did not explode. It appeared to stop, then continued to move slowly from right to left across the valley while maintaining a very bright, stable white glow. The witness reports that this was the moment when he became frightened, because the object no longer matched his immediate expectations for a firework, drone, aircraft or normal light source. The object looked to the witness like a very close, very bright star moving through the valley. He quickly grabbed his phone, closed the windows because he was scared and unsure what the object could be, and started recording. According to the witness, the camera only partially captures what was visible to the naked eye. During recording, the object appeared to gradually slow or stop. The witness reports one blink, a reduction of brightness by approximately half a tone, then apparent upward movement with additional blinking and disappearance. He states that he saw this directly with his own eyes and that the experience felt disturbing because he could not identify the object in conventional terms. The witness also states that he is not normally involved in UFO topics and does not approach this subject from prior belief. The report was created because the event was observed directly outside his windows and because a video fragment exists that supports part of the visual observation. 

The AI Agent was given all the available information on the case and instructed to perform its analyses. Its response was:

 

Initial Observation

I notice that two pieces of evidence appear to be present:

The still image you attached (daytime landscape near Acquaviva Picena) showing a very small distant object. The video file appears to be a nighttime recording overlooking a landscape with distant lights.

That discrepancy is worth noting immediately because it may mean:

·        the still image and video are from different moments,

·        the still image is a zoom/crop from another recording,

·        or the daytime image is simply being used to indicate the object's location.

For a formal investigation, I would treat them as separate pieces of evidence until confirmed otherwise.

Technical Review of the Video

I extracted and reviewed representative frames from the uploaded video.

Video Characteristics

·        Vertical smartphone format.

·        Approximately 56 seconds in duration.

·        Significant low-light noise.

·        Heavy digital compression.

·        Distant horizon lights are visible throughout.

·        No obvious structural details visible on the object.

·        Image Quality Assessment

Evidence quality for identification:

·        Category Rating

·        Stability: Moderate

·        Resolution of target: Poor

·        Lighting: Poor

·        Range estimation: Unknown

·        Object detail: Very poor

·        Motion analysis potential: Moderate

The object occupies only a handful of pixels, meaning any apparent shape is almost certainly a camera artifact rather than a true physical outline.

Preliminary Motion Assessment

Based on the available footage:

Observed

·        Bright point-like light.

·        Appears isolated from background lights.

·        Apparently slow movement.

·        No obvious sound reported.

·        Possible disappearance or fading.

·        Not observed

·        No abrupt acceleration.

·        No right-angle turns.

·        No rapid altitude changes visible.

·        No obvious structured craft shape.

·        No visible exhaust.

·        No visible aerodynamic surfaces.

Therefore:

The video alone does not presently demonstrate extraordinary flight characteristics.

Recommended Digital Processing

If I were conducting a [higher-level] technical review, I would run:

Stage 1: Stabilization

·        Horizon lock

·        Frame registration

·        Camera shake removal

Purpose: Determine whether apparent motion belongs to the object or the camera.

Stage 2: Contrast Enhancement

Useful filters:

·        Histogram equalization

·        CLAHE

·        Gamma enhancement

·        Noise reduction

Purpose: Recover faint motion and possible blinking patterns.

Stage 3: Frame Stacking

Stack multiple frames around the object.

Purpose: Improve signal-to-noise ratio to determine if object remains point-source or resolves into structure.

Stage 4: Light Curve Analysis

Measure brightness frame-by-frame.

Purpose: Distinguish between:

·        Aircraft landing lights

·        Drone LEDs

·        Planetary scintillation

·        Reflection events

·        Unknown source

·        Possible Conventional Explanations

At this stage, these remain the leading possibilities:

1. Aircraft (Most Likely Candidate)

A distant aircraft approaching or moving obliquely can appear as:

- Brilliant white light
- Little apparent movement
- Silent
- Hover-like

Especially if landing lights are facing the observer.

Probability: Moderate to High

2. Helicopter

Possible if:

- Flying slowly.
- Viewed at distance.

Less likely if no red/green navigation lights become visible.

Probability: Moderate

3. Drone

Possible if relatively nearby.

However: The witness specifically reports that it did not resemble a typical drone.
Because no scale information is available, this remains possible but unconfirmed.

Probability: Moderate

4. Atmospheric / Astronomical Object

Possible if:

- Observation occurred near the horizon.
- Object was actually stationary.

Candidates:

·        Venus

·        Jupiter

·        Atmospheric refraction

·        Temperature inversion

Probability: Low to Moderate

5. Satellite

Less likely.

Satellites generally:

·        Move steadily.

·        Do not appear to hover.

·        Usually do not remain intensely bright for extended periods near the horizon.

Probability: Low

Unknown Category Assessment:

Current classification:
Insufficient Information (Pending Investigation)

Reason: The object is genuinely unidentified at present, but the available evidence does not yet eliminate conventional explanations.

Reconstruction Possibilities

If the object were a physical luminous object, the imagery is most consistent with:

Scenario A

Small spherical light source

·        White

·        Self-luminous

·        No visible structure

Common witness description: “white orb”

Scenario B

Distant aircraft landing light

- Appears perfectly spherical
- Extremely bright
- No visible body

This is often what witnesses report as a “white fireball” or “white orb.”

Scenario C

Drone with forward-facing LED

- Small object
- Bright light overwhelms sensor
- Physical structure hidden

Additional Data Needed:

To advance this investigation substantially, I would want:

·        Exact date

·        Exact local time

·        Direction of observation (azimuth)

·        Estimated elevation angle

·        Duration

·        Witness location coordinates

·        Original uncompressed video

·        Weather conditions

·        ADS-B / FlightRadar24 data

·        Local drone restrictions

·        Satellite pass data

·        Preliminary Conclusion

At this stage, I would rate the case as: Unidentified — Insufficient Data

The object appears to be a genuine luminous target observed by the witness and captured on video. However, the footage currently lacks enough detail to distinguish confidently between:

·        Distant aircraft lights,

·        Helicopter,

·        Drone,

·        Atmospheric/astronomical source,

·        A genuinely unknown aerial object.

The strongest conventional hypothesis at present is a distant aircraft viewed head-on or at a shallow angle, though that explanation remains unconfirmed without time, direction, and flight-traffic correlation.

Conclusions of Preliminary Test of AI Agent 

The analyses performed by the AI Agent in this preliminary test demonstrates the difficulties faced by UAP investigators in order to evaluate or explain reports received from witnesses. Even when a witness cooperates fully with investigators, there are often missing data and other information necessary to arrive at a confident conclusion. Since most UAP reports often lack important details, the challenge of evaluation cannot be overstated.

If AI Agents can simplify or streamline the investigative process, UAP researchers and investigators can find their task eased considerably. The improvement in the design of AI Agents will benefit ufology, as other applications.

Labels: Agent, AI, analysis, Canada, photograph, report, research, test, UAP, UFO, video