THE BOTTOM LINE

Semi-automated offside is a narrow measurement and alert system, not an AI referee. Its strongest case is faster, more reproducible positional evidence; its limits begin where the law requires interpretation, the tracking view is uncertain or the competition has not tested the full matchday chain.

01

The short answer: it narrows the geometry

Semi-automated offside technology, or SAOT, answers a constrained set of questions. It helps locate the relevant attacker and second-last defender, identifies the moment the ball is played, generates virtual offside lines and sends an alert to match officials. That is materially different from deciding every element of an offside offence.

The distinction matters because offside has both measurable and interpretive parts. A system can estimate that a player's playable body part is beyond the line at the kick point. It cannot, by position alone, settle whether a player interfered with play, interfered with an opponent or gained an advantage under the law. FIFA's 2026 explanation explicitly says the advanced system is limited to positional offside and does not determine interference by a player who does not touch the ball.

02

The decision chain, step by step

FIFA's certification material breaks the system into four technical components: virtual offside-line placement, ball tracking and kick-point identification, skeletal tracking, and the alert logic. The February 2026 test manual then adds a logic-and-match test for top-level systems. Reading those documents together turns the feature from a black box into a reviewable chain.

A useful way to audit any claim about an 'AI referee' is to ask which link in this chain the product actually supplies. A tracking provider may estimate body points; a connected ball may improve kick-point timing; an offside system combines those inputs; the officials then verify the incident and apply the Laws of the Game.

  • Sense: cameras and ball systems capture player, ball and pitch data.
  • Synchronise: the system identifies the relevant kick point.
  • Measure: skeletal points and pitch calibration produce attacker and defender positions.
  • Alert: logic flags an onside or offside scenario and creates a visualisation.
  • Review: match officials check the proposed kick point, players and line before applying the law.
03

What changed with advanced SAOT in 2026

At the 2022 World Cup, SAOT information was routed to the video assistant referee. FIFA said the advanced 2026 version could send clear positional-offside information directly to on-field officials, allowing an assistant referee to flag earlier. The governing body's stated objective was to reduce the delay between the offside and the flag while preserving human control for interpretive cases.

FIFA and technology partner Lenovo also introduced one-second 3D player scans for the tournament. FIFA described the resulting player avatars as an input that supports identification and tracking during fast or obstructed movement, and as a way to make broadcast recreations more recognisable. Those are FIFA and partner implementation claims, not independent proof that every match situation is error-free.

04

Primary-source evidence ledger

The strongest evidence is not a promotional demo. It is the combination of the law, the test standard and the stated operating boundary. The three sources answer different questions and should not be collapsed into a single claim of accuracy.

The comparison below is the practical evidence base for this review. It shows what is established, and what still needs competition-level verification.

  • IFAB Laws 2026/27: competitions may use technology to assist with offside, including SAOT; the framework still places decisions with the referee.
  • FIFA test criteria: certification evaluates virtual lines, ball and skeletal tracking, and alert logic as separate components.
  • FIFA February 2026 manual: Quality Pro systems face controlled line-placement tests plus tracking and logic-and-match tests coordinated by an approved independent test institute.
  • Match-test threshold: each defined scenario must produce the correct outcome and relevant visualisation; an offside alert must appear within three seconds of the receiver being visually confirmed on the provider system.
  • Deployment caveat: the manual says organisers using direct-to-official alerts should perform their own due diligence on functionality, reliability and latency before competition use.
05

Where human judgment still begins

The IFAB protocol is designed around assistance, not delegation. A referee must make a decision; the VAR has independent access to footage and can assist within defined review categories. For an offside check, technology can make the positional evidence faster to retrieve, but it does not rewrite the law or remove accountability from the officiating team.

This is clearest in interference cases. A player can be in an offside position without committing an offence. Officials may need to judge line of vision, a challenge for the ball, an obvious action affecting an opponent or a deliberate play by a defender. Those are football and law questions that cannot be inferred from a red-and-blue line alone.

  • Machine-owned evidence: tracked coordinates, candidate kick point, generated line and alert.
  • Human validation: correct frame, correct attacker and defender, credible tracking and usable replay.
  • Human judgment: interference, deliberate play and the final application of Law 11.
  • Competition ownership: configuration, fallback process, communications and post-match review.
06

The failure modes that remain visible

FIFA's own offside-technology overview identifies pitch topography, camera-angle distortion, player occlusion and individual body-part tracking as recurring sources of virtual-line error. The 2026 test manual addresses these through calibration, controlled 2D and 3D scenarios, skeletal tracking tests and a manual-line fallback. Certification reduces risk; it does not mean the operating environment stops changing.

One limitation is especially important. The February 2026 manual says the 3D line-placement test using players did not yet have a pass/fail criterion, with an error metric still being explored through research projects. That does not invalidate the wider certification programme, but it is a reason to reject blanket claims that every 3D body-point problem has been solved.

  • Occlusion: several bodies hide the relevant limb from a camera view.
  • Kick-point uncertainty: the automatically selected frame differs from the manually reviewed contact frame.
  • Calibration drift: the physical pitch model and camera model no longer align closely enough.
  • Identity error: the system or operator selects the wrong attacker or defender.
  • Latency or communications failure: a correct alert reaches officials too late or through an unreliable channel.
  • Interpretation error: accurate positional evidence is applied to the wrong legal question.
07

A deployment checklist for leagues and clubs

A competition evaluating offside technology should test the complete decision path in its own stadiums, not just accept a component certificate. The provider should demonstrate the same camera, calibration, data and communications configuration intended for matchday, along with the operator workflow and the manual fallback described in FIFA's test material.

The review should also include match officials, competition operations, broadcasters and data-protection staff. The system affects more than line accuracy: it changes when a flag may be raised, what evidence reaches the referee, what is shown to supporters and how player scans or tracking data are handled.

  • Record certification mark, tested configuration and any material matchday changes.
  • Run crowded, obstructed and unusual-body-position scenarios in every stadium type.
  • Measure end-to-end alert time, not only model inference time.
  • Rehearse loss of tracking, ball data, network connectivity and referee communications.
  • Preserve an audit trail of the kick point, selected players, line, operator actions and final decision.
  • Explain to supporters what the technology measured and what the officials decided.
08

Limitations of this review

This article is a document-based implementation review, not an independent laboratory test of a live provider. FIFA's February 2026 manual discloses the certification method, but provider-level results and match-by-match error logs are not published in the sources reviewed here. We therefore do not claim an error rate, a universal accuracy figure or a measured reduction in decision time.

FIFA is both the system steward and a principal source for the 2026 implementation; Lenovo is its technology partner. Their material establishes the intended design and deployment claims, while IFAB establishes the legal boundary. Independent, competition-level reporting would still be needed to judge real-world performance across stadiums, providers and edge cases.

  • No vendor ranking: provider test results were not available for like-for-like comparison.
  • No universal performance claim: certification criteria are not the same as a public season-long audit.
  • No automation claim beyond position: interference and final legal judgment remain human responsibilities.
  • Evidence cut-off: sources were checked on 9 October 2026.

PRIMARY & AUTHORITATIVE SOURCES

Source notes

  1. FIFA — Advanced SAOT at the FIFA World Cup 2026
  2. FIFA — Testing criteria for semi-automated offside systems
  3. FIFA — Offside Technology Test Manual, February 2026
  4. The IFAB — Laws of the Game 2026/27 changes
  5. The IFAB — Video Assistant Referee protocol
  6. FIFA — Offside Technology error sources
  7. FIFA and Lenovo — AI-powered football innovations announced in 2026

Sources are selected for direct relevance. Vendor sources establish what a product or partnership claims; they do not independently prove performance. We state that distinction in the analysis.