Nine Months of Testing X's Algorithm Found It Consistently Favors the Far Right
What happened
Sky News ran a nine-month experiment on X (formerly Twitter) using nearly 90,000 posts and a set of simulated British user accounts. The goal was to document what the platform's recommendation engine actually surfaces, not what the platform claims it surfaces. The results, published in a major investigation, showed the algorithm delivered a majority share of right-wing political content across a wide range of scenarios, including posts featuring hateful or extreme language, immigration-focused narratives, and accounts associated with figures favored by X's owner.
The most striking finding was that the bias did not depend on a user's expressed preferences. Accounts seeded with left-leaning signals were still steered toward right-wing posts at a consistent rate, with right-wing content exceeding 60 percent of recommended material in some comparisons. That pattern points to a design choice rather than a side effect of popularity. The algorithm was not simply returning what was most engaged with; it was shaping the information diet of UK users in a direction that crossed ordinary engagement-driven logic.
Three factors drove the outcome. The core parts of X's algorithm are not disclosed, so there is no independent baseline against which to test claims about how recommendations work. X has also sharply restricted academic access to its data API since Elon Musk's acquisition, which limits the ability of outside researchers to run the kind of controlled checks Sky News had to build from scratch. Platform policy shifts since the ownership change appear to have reduced the reach of left-leaning voices while amplifying right-leaning ones, suggesting that owner preferences translated directly into operational choices.
The implications run past individual users. Recommendation systems at platform scale shape what political content seems normal, which voices seem dominant, and which ideas reach broad audiences. A platform that systematically amplifies extremist framing, even to users who did not seek it out, functions as a mechanism for political distortion rather than a neutral communications channel. The Sky News investigation put numbers on that shift for the UK, but the architecture behind it applies everywhere X operates.
What the investigation could not produce is a record of what the algorithm is actually doing internally, because X does not allow that record to exist outside the company. There is no independent log of why a specific post was surfaced to a specific user, no audit trail that would let regulators or researchers verify whether a given output came from engagement signals or from something else entirely. A provable record of what a system did, preserved outside the platform's own infrastructure, is precisely what would make claims about algorithmic neutrality testable rather than asserted. Without it, the only available method of accountability is a nine-month experiment run by a television newsroom.
Reported impact
- Affected parties
- Not publicly disclosed
- Harm type
- Not publicly disclosed
- Scale
- Not publicly disclosed
- Financial impact
- Not publicly disclosed
- Regulatory action
- Not publicly disclosed
Classification
Relevant governance controls
Governance control mapping is not available for this record.
- No controls mapped
Not publicly disclosed
Control mapping is analytical. It does not state that any control would have prevented the incident.
Sources and evidence
This record was researched and written by the Index. The event is also catalogued in the following database, which is listed for cross-reference.