A couple dining on a terrace watched a dark object travel towards La Défense. Their separate descriptions, and the weather reconstructed afterwards, led GEIPAN to favour a balloon. The report is also a useful example of why an archive’s category and the wording of its reasoning should both survive a retelling.

Two accounts of one passage
The observation was made on 23 August 2018 at about 20:45. The husband submitted his questionnaire the following day; his wife subsequently supplied her own account. GEIPAN corrected a one-day discrepancy in a questionnaire field against the narratives. The report’s incident date is therefore 23 August, rather than a second encounter on the following evening.
The witnesses agreed on a straight passage from north-west towards south-east. They differed on the outline: one thought it rectangular, the other rounded or elongated. A brief red glint appeared in one account. They compared the object with drones they had seen previously but felt this looked different. No photograph accompanied the report.
The setting reconstructed
The investigation checked the sunset and available weather observations. It found that the lighting described by the couple was consistent with the time, and that the recorded wind direction agreed with the reported direction of travel. The weather material supplies a reason to consider something drifting, rather than merely an assumption based on its appearance.
The source also records a practical limit: the interval before the case was processed prevented a historical traffic check. That missing check belongs to the enquiry’s documentation. It should not be transformed into a claim that no aircraft were in the area, nor into a later claim that an aircraft was positively identified.
Why a balloon was favoured
GEIPAN considered the combination of outline, silence, straight movement and wind agreement a good fit for a balloon. It suggested the brief red appearance could have been a reflection. The report describes this as a very probable explanation while acknowledging that it could not be formally confirmed.
This leaves two useful distinctions. A source-supported explanation can be the leading assessment without the precise object being recovered or its operator traced. Equally, disagreement between two viewers about a small distant outline need not require two different objects. The accounts can be preserved without forcing either witness’s shape into a more definite catalogue tag.
Reading the category alongside the reasoning
The conclusion labels the case A. Its prose, however, retains the word probable. UKPX records Possibly Explained to keep that material qualification visible. This is an editorial mapping of the report’s reasoning into the UKPX outcome vocabulary, not a claim that GEIPAN changed its own category.
The full investigation is preferable to a short search result because it sets out the alternatives, the available checks and the limits of the identification. Readers can distinguish the witnesses’ observations from the investigator’s interpretation without losing either part of the story.
Map and illustration
The map uses the published commune reference for Houilles. It does not locate the terrace or identify a private building. There is no public-place recommendation associated with this case. The symbolic rooftop illustration does not reproduce a witness photograph or attempt to decide which outline was correct.
The incident belongs to 2018; the downloadable report is dated 2023. UKPX reviewed it in November 2026. These dates describe different stages of the record and are not interchangeable.
Sources and review
Sources checked 21 November 2026. Official GEIPAN investigation, 5 October 2023, 13 pages · Broad Houilles coordinate reference

