Skip to main content
NOPE
Low reported severity Media Coverage

Nova Scotia woman says duct cleaners showed her an "after" photo of work she found undone; examiners for Global News found an AI watermark on it

A woman in Dartmouth, Nova Scotia, told Global News that a duct-cleaning crew she booked through a Facebook Marketplace advert came to her home on 9 October 2026, showed her a phone photo of a dirty furnace motor, raised the price to $300 and then showed her a photo of a motor that looked almost new as proof the job was done. When she checked after the crew left, she says, the motor was still dirty. An AI researcher at the University at Buffalo and two students, examining the photos for Global News, found a Google SynthID watermark on the clean-motor photo; the researcher said the AI detector Sightengine gave it a 99 per cent AI score and called the result their strongest evidence that the image may come from AI. She says the company denied wrongdoing and refunded her only after repeated complaints; Halifax police told her it was a civil matter.

AI system
Unidentified image tool (suspected)
Occurred
9 Oct 2026
Reported
22 October 2026
Event location
Canada
What the AI did
Communicated with the person
Reported harm
Other Material HarmPsychological Distress
Whose AI use
Someone else’s AI use
Setting
Everyday life · Finance
Evidence
AI involvement suspected · Causal attribution alleged · 1 source
5 claims: 5 reported. 4 open questions
People reported harmed
1 person

AI system as recorded: An image-generation tool, not named, that examiners commissioned by Global News say probably produced the "after" photo of a cleaned furnace motor shown to the customer; they found a Google SynthID watermark on it

What Happened

The visit. According to the customer's account to Global News, she answered a Facebook Marketplace advert for home duct cleaning at $150, arranged the visit over WhatsApp, and two men came to her home on 9 October 2026. She says a worker opened the furnace, showed her a photo of a very dirty motor and told her she had mould; the workers proposed an extra charge and the parties agreed on $300 for the job. At the end, she says, the worker showed her a photo on his phone of a motor that looked almost new, and she was impressed.

What she found. After the crew left she inspected the furnace herself and, she says, it was still filthy, with the dirt still in the section where the filter goes. The company later sent her the photos. She says she spent hours messaging the company, which denied any wrongdoing, and that she was refunded the $300 after repeated complaints.

The examination. Siwei Lyu, an AI researcher at the University at Buffalo, and two of his students analysed the photos for Global News. Their report found a Google SynthID watermark on the photo of the clean blower motor, and Lyu said the result was their strongest evidence that the image may come from AI. He added that photos on the company's website also appear to be AI-generated, while its customer review videos showed no signs of AI use. Global News' photo captions differ on which of the two photos is believed to be AI-generated; the watermark finding concerns the clean-motor photo.

Responses. Global News emailed and phoned the company and received no response. Halifax Regional Police told the customer her case was a civil matter, and a police spokesperson confirmed to Global News that it is not considered fraud.

Reported harm

The customer says she paid $300 for a cleaning she found had not been done after being shown a photo of a clean motor, felt scammed, and got the money back only after hours of complaints (Global News).

Outcome

Resolved

According to Global News (22 October 2026), the company denied wrongdoing and refunded the $300 after the customer complained repeatedly. Halifax Regional Police told her the matter was civil, and a police spokesperson confirmed to Global News that the case is not considered fraud. The company did not answer Global News.

What remains unknown

  • Whether the "before" photo of the dirty motor was also generated; Global News' captions differ on which photo is believed to be AI-generated.
  • Which tool made the photo and who in the business produced it.
  • When the customer paid relative to being shown the "after" photo.
  • Whether other customers of the same business were shown generated photos.

What the evidence supports

AI involvement: suspected. The business presented the customer with a photo of a cleaned furnace motor as proof that the job was done (communicated_with: AI-generated content an operator presented as information that a described person relied on, followed by an adverse consequence). Examiners commissioned by Global News (a University at Buffalo AI researcher and two students) found a Google SynthID watermark on that photo, and the researcher reported a 99 per cent AI score from the detector Sightengine and called the result their strongest evidence that the image may come from AI, so the AI generation rests on a named examination and is stated as probable. The customer says she was impressed by the photo and then found the work undone. The tool used is not named.

5 claims: 5 reported. What the statuses mean

Reported The customer says she booked a $150 duct-cleaning service advertised on Facebook Marketplace, two men came to her home on 9 October 2026, a worker showed her a photo of a very dirty motor and said she had mould, and the parties agreed on $300 for the job.

Causal attribution. The customer's account to Global News.

  • globalnews.ca(opens in new tab) supports · English
    'she saw an ad on Facebook Marketplace promoting home duct-cleaning services for $150' ... 'two men showed up at her home on Oct. 9.' ... 'They eventually agreed on $300 for the entire job.'
Reported The customer says that at the end of the visit the worker showed her a photo on his phone of a motor that looked almost new, and that when she checked after the crew left the motor was still dirty.

Causal attribution. The customer's account; her own photo of the uncleaned motor is described by Global News but its analysis is not reported.

  • globalnews.ca(opens in new tab) supports · English
    'this motor that looks almost new. I was impressed.' ... 'she was shocked to see a completely different result from the one she was shown in the photo.' ... 'It was still filthy. All the dirt was still in that section where the filter goes.'
Reported An AI researcher at the University at Buffalo and two of his students, examining the photos for Global News, found a Google SynthID watermark on the photo of the clean blower motor; the researcher says the AI detector Sightengine gave the photo a 99 per cent AI score and called the result their strongest evidence that the image may come from AI.

Causal attribution. Analysis commissioned by Global News; the researcher states the result as probable, not certain.

  • globalnews.ca(opens in new tab) supports · English
    'He and two of his students created an AI report for Global News, which found a Google synth ID watermark on the photo of the clean blower motor, which is a way of identifying AI-generated images.' ... 'Lyu says AI-detector Sightengine also gave the photo a 99 per cent AI score.' ... 'So that’s our strongest evidence this image may come from AI,'
Reported The customer says the company denied wrongdoing and refunded the $300 only after she complained repeatedly; Global News received no response from the company.

Causal attribution. The customer's account and the outlet's own inquiry.

  • globalnews.ca(opens in new tab) supports · English
    'She says she spent hours messaging back and forth with the company, which denied any wrongdoing. After repeatedly complaining, she was refunded the $300 she had paid.' ... 'Global News sent an email and called three phone numbers associated with the business, but received no response.'
Reported Halifax Regional Police told the customer the matter was civil, and a police spokesperson confirmed to Global News that the case is not considered fraud.

Causal attribution. Police response as reported by Global News.

  • globalnews.ca(opens in new tab) supports · English
    'she was told by police that her case was considered a civil matter and not fraud.' ... 'A police spokesperson confirmed to Global News the case is not considered fraud.'

Sources

1 source inspected. Sources that repeat one account do not corroborate each other.

How the sources were read, and where the events happened

Read in English on 2026-10-24 from the Global News page of 22 October 2026 (HTTP 200). Applies to s1.

Event countries: Canada. Affected-person countries: Canada. Court countries: Unknown.

Global News reports that the crew came to the customer's home in Dartmouth, Nova Scotia, and that she spoke to Halifax Regional Police.

Reviewed for publication 2026-10-24: Published as a core case (communicated_with): a customer says a duct-cleaning crew showed her a photo of a cleaned furnace motor as proof of work she then found undone, and examiners commissioned by Global News found an AI watermark on that photo. The AI origin is stated as probable, the company denied wrongdoing and did not answer Global News, and police treated the matter as civil.

People described

An adult customer in Dartmouth, Nova Scotia, who booked a home duct-cleaning service advertised on Facebook Marketplace and says she was shown a photo of cleaning work that she then found had not been done.

People reported harmed in this case

1 person

0 AI participants · 1 other person harmed

One customer, interviewed by Global News, who says she was shown the photo and paid for work not done. Other customers of the same business are not reported as harmed and are not counted.

Counted once within this case. The same person may appear in other cases. This count does not establish AI causation.

Tags

#suspected-ai-image#home_services#consumer-dispute#synthid#canada

Cite this case

Compiled per our published methodology: verification statuses, sourcing standards, and corrections process.

APA

NOPE. (2026). Nova Scotia woman says duct cleaners showed her an "after" photo of work she found undone; examiners for Global News found an AI watermark on it. AI incidents. https://nope.net/incidents/2026-nova-scotia-duct-cleaning-customer-shown-ai-watermarked-after-photo-of-work-found-not-done

BibTeX

@misc{2026_nova_scotia_duct_cleaning_customer_shown_ai_watermarked_after_photo_of_work_found_not_done,
  title = {Nova Scotia woman says duct cleaners showed her an "after" photo of work she found undone; examiners for Global News found an AI watermark on it},
  author = {NOPE},
  year = {2026},
  howpublished = {AI incidents},
  url = {https://nope.net/incidents/2026-nova-scotia-duct-cleaning-customer-shown-ai-watermarked-after-photo-of-work-found-not-done}
}

Related cases

Low Unidentified image tool

Stanford University: the Residential & Dining Enterprises department used generative AI to alter a promotional photograph of three students, replacing a Hispanic male student with an AI-generated Black woman, slimming two students' faces and changing their clothing, on campus banners; the university said the undisclosed alteration violated its AI policy, apologised and opened an investigation (Stanford Review, Stanford Daily, NBC Bay Area, NBC News, 21-24 September 2026)

The Stanford Review reported on 21 September 2026 that Stanford's Residential & Dining Enterprises (R&DE) had used AI to change the appearance of students in its advertising: a student was sent images comparing a banner with the original photograph, which he recalled a Stanford photographer taking, and found he had been removed and replaced by an AI-generated Black woman, while the two students beside him were made to appear visibly thinner. Stanford confirmed to The Stanford Daily on 23 September that R&DE used generative AI to modify the appearance of several students in the image; the Daily reported that the students' clothing was changed into Stanford merchandise, two students' faces were slimmed and one student's race, gender and appearance were entirely changed, and the university said the alteration and its non-disclosure violated its policy prohibiting AI in producing or altering images of Stanford people; the banner at the Governor's Corner housing centre was taken down and staff training promised. On 24 September NBC News reported the university's statement that the alteration was 'a serious error in judgment', that it had apologised to the students whose images were altered or erased, and that it had opened an investigation. The replaced student, who first found the edit funny, said that seeing his identity changed and being left out made him feel 'silenced and erased from a representation that was supposed to include me', that it was upsetting, and that he was exhausted by the national media attention. According to the student, the original photograph was taken at a 2024 Lunar New Year dinner in a campus dining hall and had been used unaltered in earlier promotional material.

High Unidentified image tool

Sonbhadra (Uttar Pradesh): police say a retired official was pressured into transferring about Rs 2.36 crore over one to one-and-a-half months by extortionists who posed as judicial and government officials and used AI-made fake nude images to threaten him; three arrested after a bank manager's alert (September 2026)

Sonbhadra police say a retired official in Uttar Pradesh transferred about Rs 2.36 crore into accounts named by extortionists, including money raised through a loan, before a State Bank of India manager noticed the repeated transfers on 22 September 2026 and alerted the cyber crime police station. At a press briefing on 25 September the superintendent of police said the fraudsters had frightened the man by invoking the higher judiciary and senior government officials (a so-called digital arrest) and had used AI to create fake obscene or nude images to blackmail him with threats of defamation and harm to his family (Times of India, Amar Ujala, ETV Bharat, Newstrack). Police froze about Rs 75 lakh, registered an FIR and arrested three people, including the alleged organiser; the outlets relay one police account, and they describe the AI element differently (images of the victim, or images sent to him during a WhatsApp chat with someone posing as a woman).

Low United Airlines chatbot

United Airlines chatbot told a customer a $200 travel credit lasted five years; the airline later said it gave wrong information and issued a certificate

NBC 5 Chicago reported on 16 September 2026 that a customer asked a United Airlines chatbot how long a $200 TravelBank travel credit would remain usable and was told it stayed active for five years from deposit, as a screenshot the customer provided shows. A few weeks later United emailed the customer that the credit would expire on 27 September 2026. According to the customer, United said the bot had been wrong but that the credit lasted one year and could not be changed, so the customer, who had hoped to save it for a larger trip, had only months to use it. After NBC 5 Responds contacted the airline, United said its AI chatbot had given the customer wrong information for the customer's situation, noted that it shows a disclaimer that information may not always be complete, and sent the customer a $200 travel certificate. The report does not say when the chatbot exchange took place.

Low Unidentified image tool

Wichita, Kansas resident reports a text with AI-generated nude images and a threat to send them to Facebook friends and family unless money was paid

KWCH (Wichita, Kansas) reported on 10 March 2026 that a Wichita resident received a message from an unknown number containing nude images that the station and the resident described as AI-generated fakes, together with screenshots of the resident's Facebook profile, friend list and family, and a threat to send the images to those people unless money was paid. The resident said the images could ruin relationships and career if they reached family or an employer, filed a police report and said no more messages arrived after blocking the numbers. KWCH reported that the Wichita Police Department said it had not seen a situation like this before. The account rests on one KWCH article built on the resident's own statements. The images were not independently examined, the sum demanded and whether any payment was made are not stated, and the date the message arrived is not stated.

If you or someone you know is struggling, free and confidential support is available. Find a helpline near you at Signpost.

Back to all incidents