Forty-tworecommendations.The press printed the twoabout Muslims and Asians.

Between July 2023 and April 2025, South Staffordshire Water led a research project called Water Efficiency in Faith and Diverse Communities, with the University of Cambridge’s Faculty of Divinity, Cambridge Central Mosque, Hindu Climate Action, the Eco Dharma Network, EcoSikh, Ecojudaism, Waterwise and five other water companies. Ofwat’s Innovation Fund, which is financed through water bills, paid £270,000 of the £300,000 budget. It was not a government campaign, and Ofwat did not run it.

The project engaged eleven communities: Muslim, Jewish, Sikh, Hindu, Buddhist, Christian and Catholic, Pagan, South and Southeast Asian households, West African households, Middle Eastern households, and a multi-faith panel of 17 to 19 year olds. Its four published documents run to 163 pages. We read all of them.

Across those documents the project makes 42 distinct recommendations. Nine concern Muslim practice. Six concern South Asian cooking. The other 27 concern Hindu, Sikh, Jewish, Christian and Buddhist practice, West African households, the general population, and, in the largest single block, the water companies themselves: build your own religious literacy, avoid othering the communities you serve, listen rather than lecture, diversify your workforce.

We counted every news article we could find on the project across two coverage waves, one at the end of July 2026 and one from 20 August, when a Telegraph Money piece set off the main cycle. Ten distinct articles, from London to Brussels to Mumbai. Every one of the ten leads with the rice recommendation, the wudu recommendation, or both. Not one headline, and not one article body we could read, mentions that the project also worked with Jewish, Sikh, Hindu, Buddhist or Christian communities. The only pages on the public internet that name those communities are the project’s own.

But the ten are not one population, and treating them as one would repeat the trick this piece is about. Three UK outlets built the frame: the Telegraph, GB News and the Daily Star. Four minority-serving outlets, Middle East Eye, Eastern Eye, 5Pillars and a Belgian Islamic centre, covered the targeting of their own readers inside a frame others had already set. Three Indian outlets carried it abroad. When a Muslim outlet puts Muslims in a headline about Muslims being singled out, it is reporting the news it owes its readers. When a national newspaper crops an eleven-community research project to the two communities its comment pages already treat as a problem, it is making the news it wants. The acts are not the same, and the census below labels which is which.

The tell is what the frame-setters did not print. The project’s documents hand any accountability reporter two real stories: a £270,000 behaviour study funded from water bills while the network loses more than a billion litres a day to leaks, and the report’s own concession that its savings evidence carried high uncertainty. In the frame-setting coverage we could read, neither is pursued. The criticism on offer is one borrowed word from a politician, “bizarre”, with the leak figure as scenery. The report itself was never the target. If it had been, the £270,000 would be in the headline. The communities are in the headline, because the communities were the point: a multi-faith infrastructure project was useful to these outlets only in the crop that fits the story they already tell about Muslims and Asians.

Watch the story move toward the narrative, step by step. The Telegraph’s published headline said “Asian communities” and dropped the feet claim that its own URL slug, “asians-told-wash-rice-and-feet-less”, and its social copy carried. GB News restored the feet and rewrote the communities as Muslim: “Muslim communities told to wash rice and feet less by Ofwat”. The Daily Star kept both. No project document tells anyone to wash their feet less. The idea appears once, in the Cambridge academic study, as a community member’s own reflection about wiping over socks, recorded alongside the note that such options could not be mandated. So the single invented fact in the whole cycle is the one that made the story most specifically about Muslim worship, and each retelling moved the headline closer to the outlet’s standing narrative. Fabrication does not drift at random. It travels toward the story it serves.

Every frame-setting headline also runs on the same verb: told. Muslims and Asians “told” to change how they wash and cook, by a regulator. The verb inverts the authorship. The wudu campaign was co-created and fronted by Cambridge Central Mosque and its Head Imam, framed as reviving the Prophet’s own practice of making wudu with about one litre, and launched during Ramadan; every participant surveyed afterwards said they would now make wudu the Prophetic way. A community teaching its own tradition, in its own voice, became a state instruction to a suspect minority, because an instruction is what the narrative needed. The most accurate headline in the census belongs to 5Pillars, a Muslim outlet: “UK water regulator-funded project urges Muslims to use less water during wudu”. Right actor, right verb. The frame’s pull still shows at the edges of the responsive coverage, and the census records it: Middle East Eye’s headline reached past its own reporting and named “the UK government” where its article body correctly named Ofwat and the water companies. That is what a frame does once it is set: even coverage pushing against it inherits its shape. It is not the same act as setting it.

The people who ran this project wrote down, repeatedly and in advance, exactly what the press then did to it.

The final report even instructs water companies to prepare a plan against misinformation, recording that during the pilot an influential figure falsely claimed the wudu campaign was a covert push to force water metering.

Two criticisms of the project survive contact with the documents, and we make both. The water network loses more than a billion litres a day to leaks while billpayers funded behaviour research; the shadow Cabinet Office minister Mike Wood called the scheme “bizarre”, and the proportion question is fair. And the project’s own final report concedes its savings evidence carried high uncertainty and calls for control groups and larger samples. The outlets that set the frame used neither, and that is the finding: coverage that wanted to hold the project to account had the ammunition on the first page of its own final report and left it there. The coverage was never too hard on the project. It never engaged the project at all. It took the two communities and left the report behind.

On 19 August 2026 the Home Office published a collection called Resources for asylum seekers in the UK: a nine-page guide and twelve further items. Two of the thirteen are about sexual consent. The rest cover housing, child safety, harassment, photography, gender equality and domestic abuse, including three addressed to asylum seekers as potential victims, with helplines. The guide states that the rules “are not just for people seeking asylum, they apply to everyone living here”. It cites no offending statistics, and the reported trigger was narrower than the coverage assumed: offenders had claimed in court not to understand UK law, and the Home Office wanted that defence closed.

Our monitoring corpus ingested 25 articles about the guide between 19 and 23 August, across ten outlets on three continents. We read every stored body in full. Of the 25 headlines, 22 frame the story through rape, sex or consent. Of the 24 bodies, 3 quote the guide’s applies-to-everyone line. 6 note that the collection contains resources for asylum seekers who are themselves victims. And 0 cite any statistic on offending by asylum seekers or migrants. Not one. The numbers that do appear are claim volumes, deportation totals and hotel counts. The wave asserts a group-specific danger 25 times and never once measures it.

GB News published 11 articles in five days, more than a quarter of its migration output that week. None of the 11 quotes the applies-to-everyone line, none notes the victim resources, and none cites a statistic. The nearest the corpus comes to the evidence question is a Labour MP’s aside on GB News that “we should probably tell a large number of men in the UK those facts as well”, and one quote in the Daily Mail, from the Green peer Jenny Jones: “There is no evidence violence against women and girls is higher amongst migrants and people seeking asylum than among the general population.” One sentence, in 24 article bodies, addressing whether the premise of the entire cycle is true.

The coverage did not need statistics because the numbers were assumed to be common knowledge. Here is what happened to the loudest of them when someone checked.

There is a reason every one of those claims had to be built from proxies. The United Kingdom does not record whether an offender is an asylum seeker. Anywhere.

Hold the two halves of that together. The government aimed a conduct intervention at a cohort whose sexual offending no official dataset can measure, while its own analysts wrote that the available data could mislead the public, and while its Ministry of Justice appeals a legal order to publish the nearest proxy the public could use to test the premise. A government that believed the numbers were on its side would publish them. This one is in court to keep them unpublished, and its answer to the question it will not let anyone measure is a poster.

The consent knowledge the posters teach is not a migrant gap on the evidence that exists. Jenny Jones, opposing the posters, cited Crown Prosecution Service research that around a third of the general population misunderstands consent within relationships and marriage. The knowledge gap is real. It is everybody’s. Only asylum seekers got the poster, and only asylum seekers were told that misbehaviour costs them their housing and their claim.

“Unfounded” in this piece means asserted without statistical foundation. It does not mean disproven, and we are not claiming asylum seekers offend at the same or lower rates than the general population. Nobody can claim that either, for the same reason nobody can claim the reverse: the data do not exist. The honest record is mixed, and here it is.

On the official measures that do exist for the wider category of foreign nationals, the picture sits close to population share, with real signals on both sides. Foreign nationals were 14.2% of sexual offence convictions in 2024 on Ministry of Justice data released under freedom of information, against roughly 13% of the adult population, on nationality fields the MoJ itself calls unreliable. They are slightly under-represented among prisoners serving sexual offence sentences, a figure deflated by roughly ten thousand deportations since 2024. They are heavily over-represented among those held on remand for sexual offences, a figure inflated by bail disparities. Sky News’s corrected estimate for Afghan-born people was around three times the UK-born conviction rate, unadjusted for age and sex. None of these figures measures asylum seekers, who are a small and distinct subset of a category dominated by long-resident European workers.

And the asylum cohort is around three quarters young adult men, while people imprisoned for sexual offences are 99% male. A cohort with that composition will show elevated raw per-head numbers even if every individual in it behaves identically to the population around them. That is not a defence of the posters. It is the arithmetic that makes honest comparison impossible without the age-standardised data the state refuses to produce, and it is why every raw figure in this debate, in either direction, is unusable as a claim about behaviour.

Individual crimes by asylum seekers are real, and we name the one that started this cycle: Hadush Kebatu, an Ethiopian asylum seeker housed at the Bell Hotel in Epping, was convicted in September 2025 of sexually assaulting a 14-year-old girl and a woman. A crime is a fact about a person. It carries no information about a rate, and a government that wanted the rate known would publish the data it is in court to withhold.

A water project studied eleven communities and made 42 recommendations, most of them aimed at everyone or at the water industry itself. Three national outlets cropped it to the two communities their standing narrative needs, printed an instruction nobody wrote, and never engaged the report they claimed to find bizarre; the minority press was left to cover its readers inside that frame, and the crop travelled the world. A government published thirteen resources for asylum seekers, most of them about housing, victims’ rights and child safety. The same week, many of the same outlets printed the rape theme 22 times in 25 headlines and measured nothing, because the measurement does not exist and the state is in court to keep its nearest proxy unpublished.

In both cases the material that falsifies the frame was one click away, in the source documents, and in both cases carrying it would have cost the story its premise. The singling out is not in the water report, which warned against it in writing. It is not established by the offending data, which nobody is allowed to see. It happens in the headline, where a register of 42 becomes a register of two, and a booklet about housing and victims’ helplines becomes a rape guide. The mechanism is opportunism: an institution publishes something crop-able, and the outlets whose narrative the crop feeds do the cropping, whatever the document actually says. That is the machine. We counted its output so you can check ours.

The recommendation register was compiled from the project’s four published documents, read in full: the South Staffordshire Water final report, the Cambridge Interfaith Programme study, the Being with Water OTHERWISE conference report and the Waterwise toolkit, all linked below. The water article census was assembled through repeated web searches in several languages and verified against live pages on 24 August 2026; where a publisher blocks automated reading, the verification note on the row says exactly what was confirmed and how. The asylum corpus is every article about the guide ingested by NewsCord’s monitoring pipeline from 19 to 23 August 2026; each stored body was read in full and tagged by hand. The corpus is a defined sample, not the universe of coverage: paywalled outlets our pipeline does not ingest, including The Times and The Telegraph, are absent from the asylum table for that reason. Claim checks cite Full Fact, Sky News’s data unit, BBC Verify and Dorset Police’s published statement. The Ministry of Justice appeal against the Information Commissioner’s disclosure order was live when this piece was published; if the data are released, we will update this piece and test our own argument against them.

NewsCord is regulated by IMPRESS. If you believe anything in this piece is inaccurate, write to [email protected] and we will correct the record publicly.