Witbe tests and monitors video services on the same real devices their viewers use, including smart TVs, set-top boxes, mobile phones and streaming sticks, on real networks, with video evidence for every result. The company can deploy and operate that device infrastructure wherever a service’s viewers are. At IBC2026, in Hall 1 at Stand 1.C53, Witbe will demonstrate what that platform becomes with AI built into every layer.
Watch a Test Build Itself on a Real Device
At the stand, Witbe will demonstrate AI prompts that no rigid script can survive, such as “select the latest episode of a named show” or “confirm everything on screen suits a child’s profile”, a human-like assessment that image-matching automation cannot make.
The agents absorb the disruptions that break traditional automation, including pop-ups, sleep mode, system prompts and overnight user-interface redesigns. When a major European streaming service reskinned its homepage for the launch of Superman, tests adapted at runtime, continued operating and even validated the coherence of the new visuals.
Beyond availability testing, the Agentic SDK, a Python framework, combines AI-driven steps with algorithmic steps in the same scenario, making it possible to measure performance indicators such as start-up time, buffering and mean-opinion-score-based video quality.
This hybrid approach allows teams to control how much AI is used in each test, from zero to 100 per cent per scenario, giving them direct control over token consumption and AI expenditure.
Witbe’s peer-reviewed IBC2026 paper, Agentic AI for Smart TV QA: A Production Field Study Across Multiple Platforms and Languages, documents eight months of agentic AI in production. Using the same test infrastructure, one engineer increased the number of maintained tests from 78 to 316 and completed 280,626 executions, absorbing every user-interface change without intervention.
Witbe COO Yoann Hinard will present the findings during the Streaming Technology 2 session on Saturday, 12 September, from 3.45–4.45pm.
Short-Form and Vertical Video QA, Everywhere
Short-form video is everywhere. Following the trend established by social media feeds, every major streaming platform has announced a short-form format built from its long-form catalogue during the past six months. Netflix, Prime Video, Disney+ and Peacock have all added vertical clip feeds within their apps.
Short-form feeds are also arriving on smart TVs, where content engineered for a phone must render on a large television screen. Witbe validates all of it in production.
For one of the world’s largest social media companies, Witbe monitors short-form feeds at a global scale, with 100 real mobile devices running more than 300,000 tests per month, together with dedicated feed monitoring on Samsung and TCL smart TVs, Fire TV and connected streaming devices.
Measured performance indicators include time to first video, feed-scrolling performance, video quality across rapid transitions, audio and video synchronisation, and advertisement load duration.
Advertisements Verified Wherever They Appear: on the Field, Beside the Live Feed and During the Break
Witbe now verifies advertisements far beyond the traditional advertising break, including two formats that were previously almost impossible to monitor: virtual advertising placements, or virtual signage, with sponsor graphics inserted directly into the game; and stay-live advertisements, where the action continues within a branded frame.
These are premium placements that are invisible to traditional verification because there is no break, SCTE marker or slate.
Witbe tracks the advertisements live against client-provided definitions, logging every occurrence with timecodes and video evidence. A monitoring set-up this specialised previously took days to configure, when it was possible at all. Using the Agentic SDK, it can now be deployed within hours.
During FIFA World Cup matches in Europe and North America, Witbe monitored broadcasts on more than 100 real devices across more than 30 locations, comparing regional feeds side by side.
Backend systems confirm that an advertisement was served. Only the device screen confirms that it was seen.
Closed Captions Verified as Viewers See Them
Caption failures are invisible to backend monitoring. The stream carries caption data, so the system reports success regardless of what appears on screen.
Witbe verifies captions as rendered on the device, including their presence, language, synchronisation, correspondence with spoken audio, graphic overlap, line count and garbled characters. AI-based analysis compares caption text with the audio, while deterministic checks handle timing and rendering. Every flagged issue links to the recorded video.
Monitoring That Explains Itself
Witbe will also demonstrate Smartgate AI, its newest intelligence layer. Questions asked in plain language, such as “what are the top failing tests this week?”, return answers grounded in measured results, together with inline charts.
Failing runs are compared with successful ones to identify the root cause. Degradation trends are surfaced across devices, internet service providers and regions before they become incidents, while the run report cues the video to the precise measurement point. Investigation times are reduced from minutes to seconds.
Shipping Continuously, All the Way to Amsterdam
Witbe releases new AI capabilities continuously. At this pace, the weeks leading up to Amsterdam will bring further advances.
“Our clients are seeing productivity gains they did not think were possible, and Witbe Suite 41 has only been rolling out for a few months,” said Mathieu Planche, CEO of Witbe. “AI was the missing link: it gives our tests the reasoning of a real viewer, so they keep working when everything on screen changes. And the pace is accelerating. We expect to arrive at IBC with even more to announce, so stay tuned.”
Book Your Session
Demonstrations, research and development briefings, and executive interviews can be scheduled at https://witbe.net/events/ibc-2026.
Visit https://www.witbe.net
