TikTok won’t win the election Slick campaigning is no substitute for ideas

Similarly, nor is this an AI election, at least not in the sense that AI-created deepfakes are distorting public perceptions. While “disinformation” campaigners like to warn about the dangers of audio deepfakes fooling the electorate, there is very little evidence of this taking place. This isn’t surprising, when you consider that AI is mostly used as a tool to analyse voter data and target us with more-or-less personalised content. It’s a handy gadget that makes campaigning easier, but, again, there’s little evidence to suggest that it’s swaying the election.

That said, obviously this is a data election, in the sense that most of the parties’ campaign spending is going into the selected digital content we encounter online and in our social media feeds. Some of this content is made for sites that don’t allow political advertising, such as TikTok. The rest is mostly paid digital advertising through Meta and Google, which has, so far, cost the Conservatives over £850,000, and Labour over £2 million.

But this isn’t new. In 2016, Cambridge Analytica became a bête noire for journalists and commentators seeking explanations for the Brexit and Trump votes. Accusations of sinister manipulation followed, as Cambridge Analytica’s bold claims of being able to psychologically profile each individual voter were taken at face value by people who had only just discovered something that had been happening for years.

In truth, the techniques of profiling voters, targeting them with tailored messaging, and measuring their effectiveness had already been honed in Barack Obama’s two successful election campaigns in 2008 and 2012. It was Jim Messina, the brain behind Obama’s digital campaigns, who then helped the Conservatives win the 2015 UK election by targeting, not just key constituencies, but key persuadable voters within those constituencies.

Almost a decade later, simple geography has continued to dominate targeting in the early stages of this election. Google only allows the use of postcode, age and gender for political advertising, but the Who Targets Me analysis of spending on Meta also shows that over half the parties’ total budget in the first week went on location-specific adverts. This makes sense: why target a seat where your candidate has no chance of winning? Or conversely, where your candidate is vanishingly unlikely to lose.

The Liberal Democrats, especially, have focused their digital resources on some key constituencies that look winnable from the Conservatives, with Facebook adverts that name the local candidate, shown only to voters in that constituency. It’s not that different from shoving leaflets through doors, except that it’s cheaper, doesn’t rely on volunteer legwork, and shows in real time which target voters respond. Smaller parties have also poured digital spend into target seats — notably the Greens in Bristol Central, and Reform in Great Yarmouth.

Parties can narrow down the audience further within a geographical area: by age, for example (around 10% of Meta adverts defined age-related targets). More specifically, advertisers can choose a Custom Audience, consisting of contacts already acquired through other means. If you responded to the Conservative Party’s 2021 survey on Pet Theft, for example, you may also have checked a box agreeing that the Party can use your email to contact you again about other campaigns. Around a quarter of the early campaign Meta adverts were aimed at a Custom Audience.

The disadvantage of this approach is that you will only address people already on your contact list. To reach voters who might be receptive, but who aren’t already on the list, advertisers on Meta can request a Lookalike Audience of profiles that resemble a Custom Audience. Around 20% of Meta ads in the first week of campaigning were targeted at a Lookalike Audience. The Conservatives were most likely to use Custom and Lookalike audiences early in this election campaign. More of Labour’s spending, by contrast, went through local candidates.

But who did they target? Labour’s 2023 Issues Survey makes providing contact details optional, but notes in the small print that those details may be used to keep in touch about policy development, compared with “other information we hold about you”, including what you might have said to a doorstep campaigner. It will, however, “be analysed in anonymised form” and not used to “re-contact you for electoral purposes”. In other words, they’re keeping this data not to target you, necessarily, but to improve the models they use to target others. This suggests a sophistication of targeting that, like the Conservatives’ 2015 strategy, can deliver surprising results simply by focusing campaigning energy where it has most effect.

Today’s transparency rules around political adverts mean that Meta’s Ad Library should show the parties targeting very specific groups of voters with very specific messages. Indeed, Labour’s adverts do show such targeting. One video ad tells the story of a man called Nat who waited “over 100 days for cancer treatment” — shown to 45-to-65-year-olds living in the “sniper alley” of potentially fatal disease onset. Another, teaching assistant Amanda’s explanation of “why she’s switching from Tory to Labour this time”, was seen mainly by women (traditionally more concerned about education and healthcare). Both also fit the genre of relatable individuals telling the story of their own political journey.

What of the Conservatives? When the election was called, their digital campaign had three themes: Clear Plan vs No Plan, “Starmer needs you to vote Reform (so please don’t)”, and immigration. After a week, a range of adverts attacking Keir Starmer, or asserting that Labour would make everyone £2,000-ish worse off, were added to the campaign.

But then, between 3 and 7 June, all Conservative digital campaigning stopped. There weren’t even any clips from the TV debate — despite it going well for Sunak. You might expect them to return a completely new strategy, in response to Farage’s return and the D-Day gaffe, but no: When advertising resumed on 7 June, the main messaging was the disputed £2,094 extra tax claim, and warnings against giving Labour a bigger majority by voting Reform or LibDem. There’s little sign of a Clear Plan from the Conservative online campaign team, but that’s not surprising given the disarray of the campaign overall.

And nor is it surprising given their digital form in previous elections. In 2019, Labour vastly outspent the Conservative party on digital campaigning. The Lib Dems, and third-party groups (mainly pro-EU groups) also spent far more than the Conservatives or the Brexit Party. Labour’s messages were shared widely, watched by millions — and failed to win them the election.

What mattered, it turned out, was not methodical data-gathering, the creation of engaging digital content, or mobilising supporters to share it with their friends and contacts. What mattered was offering the electorate something they wanted enough to vote for it.

In a close-run electoral race, canny use of data and ruthless targeting of swing voters in key constituencies can change a result. Microtargeting online messaging is a cheap and efficient way to do that, especially in an age when most of us spend so much of our time online. But to focus on the technology is to miss the bigger picture. Just as Cambridge Analytica did not explain Brexit or Trump, and ruthlessly efficient use of voter data didn’t explain Obama’s two terms, in 2024 it will not be TikTok Wot Won It. Not even for Nigel Farage.

---

Timandra Harkness presents the BBC Radio 4 series, FutureProofing and How To Disagree. Her book, Technology is Not the Problem, is published by Harper Collins.

TimandraHarknes