How we say who is leading
We gather every national presidential poll that Kenyan research firms publish, check each one, and count who comes first, second and third in each. Nothing is guessed and nothing is weighted by us.
The answer today, step by step
These are the 6 polls with fieldwork in the last 6 months (since 7 April 2026). For each one, the top three (people with the same share share a place):
| Poll | Top three | Disclosure |
|---|---|---|
| Politrack Africa15–25 Sep 2026 | Ruto 45.1% Sifuna 19.6% Kalonzo 10.3% | Partly disclosed |
| Mizani Africapublished 25 Sep 2026 | Ruto 38.7% Kalonzo 27.5% Sifuna 13.2% | Method not published |
| Swiss Poll Internationalpublished 20 Aug 2026 | Ruto 31% Kalonzo 22% | Method not published |
| Infotrak22–26 Jun 2026 | Ruto 32% Kalonzo 13% Sifuna 12% Matiang'i 12% | Fully disclosed |
| TIFA Research13–22 Jun 2026 | Ruto 24% Sifuna 15% Matiang'i 14% | Fully disclosed |
| TIFA Research2–11 May 2026 | Ruto 24% Kalonzo 19% Matiang'i 14% | Fully disclosed |
Counting the places across all of them:
| Rank | Name | Led | 2nd | 3rd |
|---|---|---|---|---|
| 1 | William Ruto | 6 | 0 | 0 |
| 2 | Kalonzo Musyoka | 0 | 4 | 1 |
| 3 | Edwin Sifuna | 0 | 2 | 2 |
| 4 | Fred Matiang'i | 0 | 0 | 3 |
So: Ruto leads all 6 polls. Second place is contested: Kalonzo is second in 4 polls, Sifuna is second in 2 polls. No one is above the 50% needed to win in the first round in any poll.
The ranking rule
People are ranked by how many polls they lead. If two people lead the same number, the one with more second places goes higher, then more third places. People with identical counts share a rank. When two candidates have exactly the same share in a poll, they share that place.
We count places rather than averaging the percentages on purpose. The firms ask different questions in different ways (see below), so their numbers are not on the same scale and an average would hide that. Whether someone comes first is something every poll can answer.
Only polls with fieldwork in the last 6 months count, because opinion moves. Older polls stay on the page under “Earlier polls” for the record.
What counts as a poll
- National. A sample of the whole country, not one county, region or party's supporters.
- All candidates. The question asks who people would vote for (or want to win) among all contenders. Head-to-head match-ups, “best opposition candidate” questions and ticket questions are left out.
- From a research firm, reported in the open. Released by the pollster or reported by an established news outlet, with a link we can show you. A screenshot shared on social media is not enough: fake polls circulate, such as one falsely credited to Infotrak that Africa Check exposed.
- Not a click-to-vote poll. Online polls where anyone can vote, including our own quiz results, only show who chose to take part.
When news reports and the pollster's own report disagree, the pollster's report wins, and we note the difference on the poll. When outlets disagree with each other and there is no report from the pollster, we use the figures most outlets agree on and say so.
How polls are found
A small program checks the news feeds of Nation, The Standard, People Daily, Capital FM, Kenyans.co.ke, K24, Kahawatungu, Eastleigh Voice, Nairobi Wire and TIFA Research several times a day. It reads stories about polls and pulls out the pollster, each candidate's share, the sample size, the method and the dates, using fixed rules rather than AI, so every number can be traced back to a sentence in a story.
A new poll goes on this page by itself only if the firm is a known pollster, at least two outlets report it with the same figures, and the numbers pass basic checks (for example, they cannot add up to more than 100%). Anything else waits for a person to check it against the sources first. The program obeys each site's rules for automated readers and reads only public pages.
The disclosure grade
Each poll carries a grade based only on what the pollster made public, not on whether we like the result:
- Fully disclosed: sample size, method, margin of error and fieldwork dates all published, interviews done face-to-face or by phone, and full results released.
- Partly disclosed: sample size and method published plus most of the rest, or the poll was done by text message or online, which reaches only people with phones or data and who choose to reply.
- Method not published: no sample size or no method. These still count, because they are real releases, but treat them with more caution.
Kenya's Publication of Electoral Opinion Polls Act requires these details to be published with a poll during the election period. Until then, many firms leave them out.
Why the polls disagree
- Different questions. TIFA asks whom you would like to win, “whether or not you have ever voted or intend to vote”. Others ask who you would vote for if the election were today. These measure slightly different things.
- Different methods. Face-to-face interviews reach people without phones; telephone and text-message surveys are cheaper and faster but reach a different mix of people.
- Undecided voters. Between 5% and 20% say they have not decided, depending on the poll. How hard a pollster pushes for an answer changes everyone's share.
- The opposition has not picked a candidate. Its support is split several ways, so second place moves around more than first.
Leading a poll is not winning
To win outright, a presidential candidate needs more than half of all valid votes and at least a quarter of the votes in at least 24 of the 47 counties. Otherwise the top two face a run-off. The election is on 10 August 2027. A lead in today's polls says who is ahead now, not who will win.
What is not here, and why
TIFA's November 2025 survey is not listed: the figures reported for it disagree with each other and we could not confirm them from TIFA's own report. Swiss Poll International published only its top two national figures, so only those are shown. We will add polls we have missed as soon as we can check them. Tell us about one.