Published 2026-07-13 by destaque.ai, a Generative Engine Optimization consultancy in Lisbon. Ler em português.
Method
We measured how consistently 84 anonymised Portuguese consumer-service brands (seven sectors, twelve per sector) are recommended across seven AI assistants (ChatGPT, Claude, Gemini, Perplexity, Grok, Mistral, DeepSeek) with live web search, over 105 prompts and 2,205 responses. A brand counts as recommended when presented as a suitable option in at least two of three runs, across at least three distinct prompts.
Findings
- 54.8% of brands were recommended by zero or one of the seven engines.
- The average brand was recommended by 1.77 engines out of seven.
- Only 49.7% of recommendations carried a qualifying citation.
- Cross-engine consistency, not mere presence, is the weak point: a brand strong in one assistant is typically absent in others.
Dataset
Brand-and-engine level dataset, 588 rows, brands anonymised by sector. Reproduces every aggregate on this page.
Download the dataset (CSV). Licensed CC BY 4.0: reuse with attribution to destaque.ai.
Limitations
Convenience sample, twelve brands per sector. Live web search reduces run-to-run stability. Recommendation measures neither service quality nor conversion. A temporal snapshot, not a time series.
destaque.ai is a Generative Engine Optimization consultancy based in Lisbon, Portugal. This study is published in full in Portuguese at https://www.destaque.ai/estudo/consistencia-visibilidade-ia-servicos-portugal-2026. Methodology available on request: contacto@destaque.ai.