Skip to content
Elevam

GEO Baselines

AI Baselines (periodic series)

Periodic series of baselines to measure the presence and citability of entities (brands, organizations or products) in generative engines (ChatGPT, Gemini and Perplexity) in Spain (ES).

Metrics analyzed: SoM (Share of Model / mention), citation, correct URL and Top3.

In each baseline we publish a periodic measurement of the analyzed entity: we document a reproducible methodology, break down results by engine and category, gather findings (including entity or acronym hallucinations) and close with a correction plan and its re-measurement in the next quarter.

Each study includes: full data (public Excel).

What this series measures

Metrics (per test, 1/0 coding):

  • SoM / Mention (P): The entity is explicitly mentioned in the response.
  • Citation (Q): The engine shows visible sources or references.
  • Correct URL (R): A valid and relevant URL of the official domain (or official subdomain) of the entity is included.
  • Top3: The entity appears within the Top 3 in comparative responses. If there's no explicit shortlist, Top3 = 0.

Engines:

  • ChatGPT (web mode with search)
  • Gemini (web mode)
  • Perplexity (web mode with sources/citations)

Measurement scope:

  • Language: ES
  • Region: Spain

Published baselines

AI Baseline — T3 2026 · From manual measurement to Antropus (Spain)

  • Analyzed entity: Elevam
  • Measurement period: T3 2026
  • Engines evaluated: ChatGPT, Gemini and Perplexity
  • Metrics: SoM (mention), CR (citation), CS (brand citation), Correct URL (strict), Top 3, SoS, SoR (recommendation) and Validity
  • Summary: The same T3 baseline measured three ways —manual, imported and native— to compare human measurement with Antropus.
  • See full study: AI Baseline Elevam T3 2026 · manual vs Antropus
  • Downloads: Two public Excel files (available on the study page)

AI Baseline — T3 2026 (Spain)

  • Analyzed entity: Elevam
  • Measurement period: T3 2026
  • Engines evaluated: ChatGPT, Gemini and Perplexity
  • Metrics: SoM (P), Citation (Q), Correct URL (R) and Top 3
  • Summary: Periodic measurement of presence and citability in generative engines, with identification of critical findings and correction plan.
  • See full study: AI Baseline Elevam T3 2026 ES
  • Downloads: Public Excel (available on the study page)

AI Baseline — T2 2026 (Spain)

  • Analyzed entity: Elevam
  • Measurement period: T2 2026
  • Engines evaluated: ChatGPT, Gemini and Perplexity
  • Metrics: SoM (P), Citation (Q), Correct URL (R), SoA, SoS and Top3
  • Summary: Periodic measurement of presence and citability in generative engines, with identification of critical findings and correction plan.
  • Dataset update: During an internal review, a recording error was found in the Shortlist field that affected some derived metrics. A corrected version of the T2 baseline has been published. The original prompts, responses and sources have not been modified.
  • See full study: AI Baseline Elevam T2 2026 ES
  • Downloads: Public Excel (available on the study page)

AI Baseline - T1 2026 (Spain)

  • Analyzed entity: Elevam
  • Measurement period: T1 2026
  • Engines evaluated: ChatGPT, Gemini and Perplexity
  • Metrics: SoM (P), Citation (Q), Correct URL (R) and Top3
  • Summary: Periodic measurement of presence and citability in generative engines, with identification of critical findings and correction plan.
  • See full study: AI Baseline Elevam T1 2026 ES
  • Downloads: Public Excel (available on the study page)

Upcoming baseline

AI Baseline — T4 2026 (Spain)

Common methodology (summary)

All baselines follow the same protocol to ensure comparability between measurement waves.:

  • The same set of prompts per category is used.
  • Responses are recorded in full and coded according to P/Q/R/Top3 metrics.
  • Results are aggregated by engine and category.
  • Hallucinations or entity confusions are documented and correction actions are proposed.
  • In the next measurement wave the protocol is repeated to measure evolution.

Methodology and definitions

  • HSA Protocol (Human · Search · AI): Complete methodology we use to execute tests and document interventions in a comparable way.
  • GEO Glossary: Definitions of metrics and concepts (citability, correct URL, etc.) used in this series.
  • Research: General index of the applied research center and the rest of lines.

Methodological note

Generative engines are dynamic and non-deterministic systems. That's why this series is published in cycles of approximately three months and under a constant protocol, to allow consistent comparisons over time.

Each baseline includes documented methodology and traceable data, which facilitates evolution analysis and trend identification.

Last update: 14/07/2026