A web-readable version of our latest review. You can download a PDF version here which includes links to the data sources.
Why we looked again
We published our campaign update “The Perfect Storm” just after the May 7th elections. There we warned that, as we transition from the digital to the AI campaigning era, and a five-party system with fragmented vote share and razor thin margins, misleading election advertising has the potential to become significantly more decisive in UK élection results.
We highlighted that right now we have a once in-a-generation opportunity to actually do something about it. We reviewed the evidence we’ve given to the Representation of the People Bill (formerly known as the Elections Bill) Committee. Several amendments are now being put forward by MPs at the Report stage of the Bill and will incorporate our recommendations on how election advertising should be regulated.
The recommendations are:
- Introduce election advertising regulation to combat misleading factual claims (as recommended by the House of Lords Democracy and Digital Technologies Committee in 2020)
- Establish an independent database of election advertising to counter “dark” election advertising (ads visible only to the recipient and not open to public scrutiny)
- Make it illegal to create or distribute digital content that falsely purports to be a political candidate (or claims to be speaking for them), with intent to deceive voters. There should be a clear exemption for parody, satire or artistic expression
- Strengthen imprint rules to require all digital and printed election material to display prominently the name of the political party commissioning or promoting the ad.
You can read our recommendations in more depth here.
However, we like to substantiate our statements, which is after all one of our asks of election advertising, and to date we haven’t in any of our election reviews taken a close look at actual election results. Accordingly, as soon as Democracy Club released its post-election results data, we began digging into the numbers to see just how close the May 7th elections had been.
What we found startled us, and demonstrated how dangerous and critical the role of advertising in fine-margin elections has become.
Elections decided at the margins
2026 saw the most contests won in recent years by a small margin (<5%)
- A staggering 41.4% of wards had a winning margin of less than 5%
- 12.2% of wards were won by less than 50 votes
- 66.7% of wards were won by a margin of less than 500 votes
The “less than 1%” club
- 398 wards had winning margins of <1%
- 9 wards were decided on a winning margin of just 1 vote.
Methodology:
- “Winning margin” is defined as the gap between the last elected candidate and the first losing candidate. For example, where wards elect multiple candidates a winning margin might be calculated between the 3rd and 4th ranked candidates if 3 seats were contested.
- Marginal contests have been defined as having a winning margin of 5% or less of votes.
- All the data in the above analysis has been sourced via Democracy Club.
Many thanks to Simon Short who conducted the analysis in this report.
Why factual accuracy in election ads matters
In elections this close, misleading campaign material does not need to shift thousands of votes to matter. In some cases, it may only need to influence a handful.
That prompted us to look again at some of the paid ads and campaign materials that circulated during the campaign period. This short follow-up report is not intended to be an exhaustive analysis of advertising that ran in the run-up to the May 7th elections. Instead, it presents a sample of the questionable ads we found: unsourced or distorted polling claims, misleading graphics, selective presentation of local data, and tactical voting messages that gave voters a false impression of the real contest in their area.
The ads are from England’s local elections and drawn from the Facebook Ad Library and leaflets submitted by Democracy Club supporters. That means this report does not capture misleading ads on other social platforms or digital ads that are largely invisible to public scrutiny – including display ads served directly on websites like Yahoo!, programmatic ads (those targeted to voters across the internet), and other media such as out-of-home advertising.
From the small sample we’ve looked at we cannot prove definitively that any individual ad changed any individual result. But that is not the right test. The same claims used in a Facebook ad will often also appear in leaflets, canvassing scripts, social media posts and other campaign materials. And when the gap between the final winning candidate and the first losing candidate is sometimes only a handful of votes, that adds even greater weight to what is seriously damaging malpractice.
This matters not only because misleading ads might help one candidate over the line. In a fragmented contest, they can also take votes away from candidates lower down the vote share, distort voters’ understanding of who is really in contention, and ultimately affect who is elected. Where several parties are deploying misleading claims in the same electoral environment, the cumulative effect can be significant, even if the impact of any single advert is impossible to isolate.
The examples that follow show how this can no longer be considered a theoretical concern. As more parties compete closely, more contests are decided at the margins and campaigns fight for every vote. Misleading election advertising has become a real threat to democracy. The case for regulation requiring transparency and factual accuracy in election advertising is now urgent.
A look at our sample of ads
Chelmsford – Reform UK
A Reform leaflet from Chelmsford gave no source for its bar chart that put Farage’s party on 34% and the Conservatives and Labour on 16%. Full Fact said that, while Reform had polled at 34% with at least one pollster in the past, it had found no exact match for the Conservative and Labour figures at the same time. They also noted the bar chart was “completely out of proportion”.
Polling for Essex, the county council area that includes Chelmsford, was at odds with the bar chart. JL Partners’ March 31st – April 13th Local Elections MRP put Reform UK on 37.8% in Essex, but also showed the Conservatives on 21.1% and the Liberal Democrats on 19.0%. The leaflet’s chart did not mention the Liberal Democrats at all, despite them being projected close to the Conservatives.
This was material because the Chelmsford-area county division results were close. Reform UK won four of the eight Chelmsford county divisions declared on 8th May, the Liberal Democrats won three and the Conservatives won one; a ninth Chelmsford division, Springfield, was postponed until 18 June. Three contests were particularly tight: Great Baddow and Galleywood, where Reform beat the Liberal Democrats by 34 votes (0.5% vote share); Broomfield & Writtle, where the Conservatives beat Reform by 286 votes (3.6% vote share); and Chelmer, where Reform beat the Liberal Democrats by 400 votes (5.8% vote share). In such close races, omitting a locally competitive party from a polling graphic gives voters a distorted impression of the contest.
We can’t say the ad was responsible for any of these results but neither can its owners say it wasn’t. Perhaps they would care to explain its misrepresentation of the situation.
Essex council polling
(Source: JL partners 2026 Local Elections MRP: March 31st – April 13th 2026 polling)
Reform UK: 37.8%
Conservative: 21.1%
Liberal Democrat: 19.0%
Labour: 14.1%
Green Party: 5.8%
Other: 2.2%
Result
Of the eight Chelmsford county divisions declared on 8th May, Reform UK won four, the Liberal Democrats three and the Conservatives one. A further Chelmsford division, Springfield, was postponed until 18th June following the death of a candidate. Results for three of the closest elections in Chelmsford are below.
Great Baddow and Galleywood
Reform beat the Liberal Democrats by just 34 votes (0.5% vote share)
| Candidate | Party | Votes | Elected? |
| Kevin Bellamy | Reform | 2,436 | Yes |
| Kieron Franks | Lib Dems | 2,402 | No |
| James Raven | Con. | 970 | No |
| Richard Joseph John Hyland | Ind. | 611 | No |
| Amit Shah | Green | 456 | No |
| Shirley Mason | Labour | 161 | No |
Broomfield & Writtle
The Conservatives were ahead of the Reform candidate by 286 votes (3.6% vote share).
| Candidate | Party | Votes | Elected? |
| Mike Steel | Con. | 3,025 | Yes |
| Chris Davy | Reform | 2,739 | No |
| Rachel Wheelhouse | Lib Dems | 1,093 | No |
| Kevin Terence Kuken | Green | 790 | No |
| Jessica Peacock | Labour | 320 | No |
Chelmer
Reform were ahead of The Liberal Democrat candidate by 400 votes (5.8% vote share).
| Candidate | Party | Votes | Elected? |
| Robert Paul Jerome | Reform | 2,388 | Yes |
| Smita Rajesh | Lib Dems | 1,988 | No |
| Rengarajan Subramanian | Con. | 1,740 | No |
| Mike Thompson | Green | 539 | No |
| Peter Alexander Dixon | Labour | 267 | No |
Birmingham – Conservatives
You can see other examples of this ad template for Bournville & Cotteridge and Edgbaston in the PDF version of our report.
The West Midlands Conservatives ran an ad template across multiple wards in Birmingham framing the contests as a Labour–Conservative fight and explicitly told voters that “only the local Conservatives can beat Labour”.
The eventual results show, however, that this framing obscured the strength of the Green vote in these areas. In Harborne, the Green Party finished ahead of the Conservative candidate by 52 votes (0.4% vote share). In Bournville & Cotteridge, Labour beat a Green candidate by just three votes (0.02% vote share), while the Conservative candidates were far behind in eighth and tenth place. Even in Edgbaston, where the Conservatives won, the strongest non-Conservative challengers were Green candidates, not Labour.
And as if that isn’t enough what was the data they were using mentioned in small print at the bottom of the ads? The election results for 2022!
This is the democratic harm at the heart of misleading “can’t win here” advertising. These claims do not just promote one party over another; they attempt to define the reality of the contest for voters. And when that reality is constructed using irrelevant data, the effect can be to steer people away from voting according to their genuine political preferences.
That is what makes this example so troubling. Voters who were sympathetic to Green candidates were effectively being told that their vote would not matter yet the actual results showed the opposite. In Harborne, Bournville & Cotteridge and Edgbaston, Green candidates were not also-rans. They were central to the real electoral contest. To present them as incapable of winning, on the basis of 2022 results, was not simply rough-and-tumble campaigning: it was a shoddy attempt to distort voters’ understanding of their democratic choices.
Result
Bournville & Cotteridge
Labour beat a Green candidate by 3 votes (0.02% vote share)
The two Conservative candidates achieved 937 and 698 votes at 8th and 10th place out of 11 candidates.
| Candidate | Party | Votes | Elected? |
| Roxanne Green | Green | 2,062 | Yes |
| Nicky Brennan | Labour | 1,738 | Yes |
| Richard George Winter | Green | 1,735 | No |
| Muhammad Ali | Labour | 1,559 | No |
| Chris Lacey | Reform | 1,369 | No |
Harborne
The Green Party beat a Conservative candidate by 52 votes (0.4% vote share)
| Candidate | Party | Votes | Elected? |
| Martin John Brooks | Ind. | 1,723 | Yes |
| Kevin James Carmody | Green | 1,463 | Yes |
| Hugo George Rasenberg | Con. | 1,411 | No |
| Richard William Moore | Labour | 1,344 | No |
| Ben Goodwin | Green | 1,289 | No |
Edgbaston
A Conservative candidate beat a Green candidate by 629 votes (7.5% vote share).
| Candidate | Party | Votes | Elected? |
| Deirdre Caroline Alden | Con. | 1,847 | Yes |
| Matt Bennett | Con. | 1,761 | Yes |
| Nathan James Hogg | Green | 1,132 | No |
| Oscar George Prosser | Green | 1,081 | No |
| Mandy Sanghera | Labour | 674 | No |
Hanwell Broadway – Ealing – Labour
A leaflet delivered by the Labour Party in the Northfields ward of Ealing Borough Council exposes a cynical manipulation of data. The material is specifically designed to falsely imply that Reform UK might win, with the explicit goal of frightening voters away from considering the Green Party, the Ealing Community Independents, or the Liberal Democrats, and instead coercing them to choose Labour.
Result
Hanwell Broadway
A Green candidate beat a Labour candidate by 151 votes (1.2% vote share)
| Candidate | Party | Votes | Elected? |
| Clare Welsby | Green | 2050 | Yes |
| Natalia Kubica | Green | 2001 | Yes |
| Andrew Walkley | Green | 1924 | Yes |
| Polly Knewstub | Labour | 1773 | No |
| Yoel Gordon | Labour | 1621 | No |
Blackheath, Lewisham – Liberal Democrats
This Blackheath ward Liberal Democrat leaflet is a clear example of a misleading “can’t win here” squeeze message. The leaflet’s headline claim was that “only the Liberal Democrats can beat Labour here” and that “the Green Party can’t win in Blackheath ward.” However, the evidence used to support this claim was not current polling, but the 2022 council election result. By presenting an old result as the basis for a contemporary tactical-voting message, the leaflet gave voters a distorted picture of the actual 2026 contest.
The May election result also showed that the claim was wrong. The Green Party won two of the three seats in Blackheath ward, with Rebecca Jones topping the poll and Tracey Martin also elected. Labour won the remaining seat. The next unelected candidate was also a Green candidate, Ben Rand, who missed out by just 89 votes (0.58% vote share). The Liberal Democrat candidates finished behind the leading Greens, meaning the election outcome directly contradicted the leaflet’s central message.
There was also public pre-election polling showing that the Greens were highly competitive in Lewisham. JL Partners/LSE and YouGov MRPs both indicated a major Green challenge across the borough, although these were borough-level estimates rather than Blackheath-specific ward polls.
The leaflet used old data to make a strong present-tense claim about who could and could not win. In reality, Greens were not only capable of winning in Blackheath; they did win, and came close to taking all three seats.
Lewisham Council polling
(Source: JL partners 2026 Local Elections MRP: March 31st – April 13th 2026 polling)
Labour: 44.5%
Green Party: 24.0%
Conservative: 12.1%
Reform UK: 8.2%
Liberal Democrat: 6.3%
Other: 4.9%
Result
Blackheath
A Green candidate beat another Green candidate by 89 votes (0.58% vote share)
Liberal Democrat candidates finished 5th, 7th and 9th in the ward vote.
| Candidate | Party | Votes | Elected? |
| Rebecca Jones | Green | 1890 | Yes |
| Pauline Dall | Labour | 1626 | Yes |
| Tracey Martin | Green | 1624 | Yes |
| Ben Rand | Green | 1535 | No |
| Chris Maines | Lib Dems | 1358 | No |
It is time to rebuild trust
The examples in this report are not isolated curiosities. They are a warning about where election campaigning is heading. While this review has focused on election ads that rely on misleading statistics, these are only one part of a much wider problem. Over the years, our reviews have documented many other examples of non-transparent, misleading and deceptive campaigning, and further evidence is set out in our report published on May 7th. These include deepfakes made of candidates, leaflets designed to appear as though they come from independent institutions with no clear mention of the political party behind them, and “dark ad” examples from the Brexit referendum.
As vote share fragments and more contests are decided by tiny margins, deceptive election advertising no longer needs to persuade thousands of people for it to matter. In some places, a handful of voters can decide who wins.
Trust in politics is already dangerously low, and election advertising is the least trusted of various forms of advertising (across banks, retailers, price comparison websites etc.)
RPA’s Opinium research found that 57% of the public mostly or completely distrust ads from political parties, while 56% said they would trust electoral ads more if they knew they were regulated by an authority such as the Advertising Standards Authority (ASA). The same research found overwhelming public support for requiring factual claims in election adverts to be accurate.
The ASA describes its ambition is to make every UK ad a responsible ad. Yet election advertising – the advertising that helps decide who holds power – remains a glaring exception: misleading factual claims in election ads are outside equivalent regulation. That gap is no longer defensible.
MPs now have an opportunity to close it and to introduce other measures discussed in this document. Amendments to the Representation of the People Bill would bring basic standards of transparency and factual accuracy into election advertising, introduce an independent database of election ads, strengthen imprint rules, and protect voters from deceptive impersonation and deepfake-style material. These are not partisan measures. They are safeguards for fair democratic competition.
Labour campaigned at the general election on restoring trust in politics. That commitment now requires leadership from the government, and support from MPs across Parliament, to bring election advertising within the basic standards expected of every other advertiser. This should concern every party. In a rapidly evolving campaigning environment, the current gap is no longer tenable. Voters already distrust political advertising more than other forms of advertising, and if misleading or deceptive campaigning is allowed to continue unchecked, the consequences will be felt not only in public confidence in elections, but by MPs and candidates in their own constituencies at the next general election.
