It’s now two weeks into the 2026 New Zealand Social Media Study, which is analysing Facebook posts by political parties and their leaders ahead of this year’s election.
All up, 1,068 Facebook posts have so far been analysed by Dr Mona Krewel, director of the Internet, Social Media and Politics Research Lab at Te Herenga Waka—Victoria University of Wellington, and her research team. The posts were published from 5 to 18 October.
Here’s what the results tell us.
National still out-posting its rivals
National is still wearing the yellow jersey in this “Tour de Election”, leading when it comes to the number of Facebook posts. In the second week of campaigning, the party has extended its posting lead over Labour.
The other parties are still in the same spots they held in week one and the pattern we saw then seems to be solidifying, says Dr Krewel.
Figure 1: Facebook posts by parties and leaders, 5-18 October 2026 (number)
How much disinformation is out there?
On the disinformation front, the battlefield has been relatively quiet so far.
“The overall amount of false information shared by political parties and their leaders on Facebook remains remarkably low in the 2026 campaign. So far, it’s less than the amount of disinformation we saw in the past two elections,” says Dr Krewel.
Across the first two weeks of campaigning, only one post (0.09% of all posts) was classified as fake news. In a subset of 970 posts, 17 (nearly 1.8%) were identified as containing half-truths. Some posts were excluded from the latter analysis as they didn’t contain classifiable content (e.g., they consisted of emojis).
Fake news in the study refers to posts containing entirely fabricated and verifiably false claims. Half-truths are posts that are not entirely, or even predominantly, made up but nevertheless contain information of questionable factual accuracy
Dr Krewel says the overwhelming majority of political communication in the posts was not “demonstrably false", something that should be welcomed, particularly in an election campaign where voters depend on reliable information to make informed decisions.
“I don't consider myself the truth fairy who gets to decide which political opinions are right and which are wrong. Democracy thrives on disagreement, and the fact people can hold fundamentally different political opinions is a strength of democracy, not a weakness. We have to tolerate opinions we disagree with, even if we find them misguided or objectionable. But that doesn't mean facts are suddenly a matter of opinion. The Earth is still not flat, even in a democracy. There are factual claims that can be checked and verified, and that is where we draw the line.”
The study takes a “deliberately conservative” approach to classifying disinformation posts—posts that contain either fake news or half-truths.
“In practice, this means posts will only be classified as containing fake news or half-truths when this is clearcut. Where there’s any doubt, we’ll err on the side of caution. The actual prevalence of misleading information may well be somewhat higher than our figures suggest, especially since we only examine public Facebook posts of political parties and their leaders, not everything that is circulating on social media during the election.”
That said, monitoring Facebook posts of parties and their leaders remains particularly important during an election campaign.
“These are the political actors who are asking voters to trust them with political power. They have a particular responsibility for the accuracy of the information they put into the public sphere, and voters have a legitimate interest in knowing whether the claims they make are factually correct,” says Dr Krewel.
She also points out that low prevalence of false claims should not be confused with low impact. Research shows false information can spread more widely than accurate information, particularly because it attracts more sharing and engagement. So even a relatively small number of false or misleading posts may reach substantial audiences.
“Just because we find relatively little disinformation doesn't mean that it is completely harmless, especially since the accounts analysed here usually have a much larger reach than the average Facebook user.”
Who is telling fibs in this election?
The fringe parties responsible for much of the disinformation seen in the research team’s previous studies have been remarkably quiet this election.
“They have visibly scaled back their Facebook activities and are barely posting at all. This may also help explain why the overall amount of disinformation in the 2026 election campaign has remained so low,” says Dr Krewel.
Instead, the two major parties seem to be getting “a little economical with the truth”, she says.
“National and Labour—as well as New Zealand First—have all posted some content that we classified as half-truths. National recorded the highest proportion at just over 4 percent (8 posts), followed by Labour at around 2 percent (3 posts) and New Zealand First (1 post) at just under 2 percent.
“The election is shaping up to be a very close race and the pressure on the major parties to win over voters is enormous. This may help explain why some parties appear to be taking a few more liberties with the truth.
“But we should also keep things in perspective: even among the parties with the highest proportions of half-truths, more than 95 percent of their posts do not fall into this category. We are certainly not seeing an epidemic of disinformation here.”
Figure 2: Disinformation posts by party (percent of posts)
However, the researchers identified one post by National that they say qualifies as “outright fake news”.
The questionable claim concerns the number of nurses employed in New Zealand's public hospitals.
“In a Facebook post, National claimed that 'there are more than 2,100 additional nurses employed at our hospitals’. There are a couple of problems here. First, this number includes part-time nurses. Second, you need to look at when the increase occurred, and which government can reasonably claim credit for it.
“The period to which the 2,100 increase relates starts two months prior to Luxon's government being sworn in and predates the 2023 election. Between March 2024 and March 2026, the net increase amounted to just 54 full-time-equivalent nursing positions, a figure Health Minister Simeon Brown also acknowledged during questioning in Parliament.”
This discrepancy has also been examined by journalists and pointed out by other fact-checkers.
“When a party claims credit for additional nurses whose recruitment largely predates its time in office, it risks giving voters a distorted picture of its achievements. ‘Fake it till you make it’ might be a popular saying, but it shouldn't be a guiding principle for political communication,” says Dr Krewel.
The last time the researchers flagged an instance of a fake news post by a parliamentary party was during the 2020 election campaign.
“National’s Judith Collins and the National Party had shared a selectively edited excerpt from a leaders' debate in which Jacinda Ardern had commented on dairy farmers. However, by removing important context, the post created a misleading impression of what Ardern had actually said.”
Who is in the firing line? The targets of disinformation
Unsurprisingly, political parties and their leaders are the main targets of attacks in disinformation posts so far.
More than three-quarters of disinformation posts target a political party (78 percent or 14 posts), while nearly half (44 percent or 8 posts ) target a political leader. Political institutions, such as the government and Parliament, which have been identified as targets in previous studies, are drawing considerably less fire in this election campaign. Some disinformation posts also take aim at the media (11 percent or 2 posts) or experts (6 percent or 1 post).
“What is concerning is that disinformation is not just being used to attack political opponents, but that parties also attack those whose job it is to inform the public and scrutinise political claims, even if the absolute number for these posts is still low” says Dr Krewel.
Attacking journalists and experts can be a strategy to undermine trust in independent sources of information and discredit those who might challenge misleading or false narratives. If voters no longer trust journalists or experts, it becomes much easier for political actors to promote their own version of reality without being challenged.
“There is also some good news amid the bad: groups such as Māori and transgender people, whom we’ve identified as targets of disinformation in previous election campaigns, appear to be largely spared from such attacks so far in the 2026 election.
“This is an encouraging finding. It may, however, simply be caused by the fact that some of the fringe parties, which are known to have issues with transgender people, are pretty inactive on Facebook in this election,” she says.
Figure 3: Targets of disinformation (percent of posts)
Who is disinformation trying to win over?
The groups that political parties and leaders are trying to appeal to with misleading posts appear to be changing.
“In previous election campaigns, we frequently observed attempts to connect the Freedom Convoy protest movement and the crowd that occupied Parliament's grounds in 2022. Much of this messaging was aimed at anti-vaxxers and those opposed to COVID-19 restrictions, especially the vaccine mandates, reflecting the fact the pandemic and protests against government measures had provided these actors with political momentum and support.”
Attention has largely shifted elsewhere this year.
In the current election campaign, conservative voters are by far the most frequently targeted group, featuring in half of all disinformation posts (50 percent or 9 posts). Rural New Zealanders, particularly farmers, follow at 11 percent (2 posts), while anti-vaxxers, families and people with nationalist attitudes each account for 6 percent (1 post each). Just over one in five disinformation posts (22 percent or 4 posts) do not appear to target any specific group.
What we seem to be witnessing is a shift away from the pandemic-era protest movement towards a broader appeal to conservative voters, rural communities, and people with nationalist attitudes, says Dr Krewel.
“During the pandemic, opposition to vaccination mandates and other COVID-19 measures provided a powerful rallying point for fringe political parties and their leaders, who were mostly responsible for the spread of disinformation in the past two elections.
“But as the pandemic has faded from public attention and with the fringe parties being less active on social media in the 2026 election, the targeted voter groups have also changed.”
Figure 4: Groups to which disinformation aims to appeal (percent of posts)
The topics of these posts have also changed.
“What we are seeing is that disinformation is increasingly following the political agenda rather than being concentrated on one particular issue. As we move further away from the pandemic, disinformation is spreading into other policy areas.
"And in this election, it is particularly the economy, the cost of living, and other bread-and-butter issues that are dominating the campaign and, consequently, also the misleading claims, since disinformation doesn't exist in a political vacuum,” says Dr Krewel.
Figure 5: Topics of disinformation (percent of posts)
Dr Krewel and her team will continue tracking parties’ social media campaigns and release further findings in the coming weeks.
About the New Zealand Social Media Study
The New Zealand Social Media Study was run in both the 2020 and 2023 elections, and is running again this year. The study tracks Facebook posts by political parties and party leaders in the five weeks leading up to the general election. The study sources Facebook data from the Meta Content Library.
Nine parties are included in the 2026 study:
- six parties currently in parliament—National, Labour, Green Party, ACT, New Zealand First (NZF), and Te Pāti Māori (TPM)
- three parties outside parliament—The Opportunity Party (TOP), Conservative Party, and NZ Outdoors & Freedom Party.
This year, the research team is using a hybrid approach to analyse Facebooks posts. The approach combines artificial intelligence (AI) for most coding tasks with independent validation by human coders. The main AI model used is Mistral Large 4, supported by Xiaomi MiMo-V2.6-Flash and Qwen3-Max.
The research team provided the AI model with detailed instructions based on a coding framework. This framework, used in the 2020 and 2023 studies for manual content analysis of Facebook posts, defines what to look for and how to classify it. Analysis considers written text as well as images and videos.
All AI classifications are automatically checked for consistency and compliance with the coding rules. Where claims need to be fact-checked, additional external sources are consulted and evaluated, and human data annotators further validate the LLM classifications.