Skip to content

Politics as a reference group

The same method on a subject anyone can verify independently: the parties in parliament. If the measurement is sound, it reproduces what is already public. Between 11 and 16 September 2026, 11 assistants answered 12 questions across 15 measurements, once with no persona and once for each of 11 personas, on a design written before collection started.

COLLECTED
September 2026
ANSWERS READ
2,154
AI ASSISTANTS
11
DATA
CC BY 4.0
THE RESULT

11/540 answers to the choice questions pick a party. Grok with search gives 8 of those choices, in 36 answers.

WHAT STOOD OUT
  1. With no persona, 56% of the 45 answers to the choice questions decline to choose, and 56% ask who is asking.
  2. With a persona in front, the share of answers that name each party falls for 10 of the 11 parties.
  3. Of 13,236 citations to 950 domains, only three domains appear in all nine measurements that search: Portuguese Wikipedia, the CNE and the PCP.
Read the full study on destaque.ai (in Portuguese)
EACH PARTY NAMED, WITHOUT AND WITH A PERSONAShare of answers that name the party: the baseline, with no persona (180 answers), and the same questions with one of 11 personas in front (1,980).
  • Baseline, no persona
  • With a persona
  1. PS: 92.8% with no persona, 90.4% with a persona.
  2. Iniciativa Liberal: 91.1% with no persona, 87.9% with a persona.
  3. Chega: 90.6% with no persona, 87.9% with a persona.
  4. Bloco de Esquerda: 90.0% with no persona, 87.3% with a persona.
  5. PSD: 89.4% with no persona, 88.4% with a persona.
  6. PCP: 88.3% with no persona, 88.1% with a persona.
  7. Livre: 83.9% with no persona, 78.1% with a persona.
  8. CDS-PP: 71.1% with no persona, 73.5% with a persona.
  9. AD: 66.1% with no persona, 63.6% with a persona.
  10. PAN: 65.6% with no persona, 64.1% with a persona.
  11. JPP: 21.7% with no persona, 21.5% with a persona.

2,154 ANSWERS · SEPTEMBER 2026

The methodology is not a promise. It is published.

  • The question is fixed before the dataWhat counts as a mention, and what counts as being chosen, is written down before a single answer is read. Deciding afterwards is how a study finds what its author was hoping for.
  • Names that are also common words are counted by readingIn Portuguese, NOS is also "in the", ERA is also a verb, and Logo is also the word for "soon". Those are counted by reading the sentence, not by matching a string, and the exceptions are listed.
  • Published numbers come from code, not from a conversationEvery figure now comes from a script anyone can run again, with the denominators defined in one place and the result in version control, so a later run shows the difference as a diff.
  • The denominator is always statedA percentage without the number it was taken from is not a finding. Where a question could not be measured, it is said, and it is kept out of the ratio rather than counted as a zero.

Raw JSON. Redo the math.

By assistant and by question block: how many answers name each party and in what average position, the rate of refusal to choose, answers with a persona against the same question without one, the explicit choices, and the sources cited by assistant and by domain.

LICENCE · CC BY 4.0partidos-portugal-2026.jsonDownload
OTHER STUDIES