62.2%60.0%
−2.2 pp
Baseline T3 2026
Elevam AI Baseline — T3 2026 (Spain)
Reproducible GEO research document (v1.0) for the T3 2026 measurement wave in Spain.
Documents the observed evolution of Elevam's presence, citability, attribution and comparative visibility in generative engines compared to the corrected T2 2026 baseline.
PublishedJuly 15, 2026
Observed state
Observed state in AI engines (T3 2026)
This study documents Elevam's observed behaviour in ChatGPT, Gemini and Perplexity during T3 2026: whether the brand appears, whether the engine displays sources, whether it cites an Elevam URL appropriate to the prompt's intent, whether the brand enters the Top 3, and what qualitative problems appear in the answers.
The comparison is made against T2 2026 using the same set of 15 prompts, the same three-group structure and the same main coding logic. The aim is not to demonstrate causality nor to automatically attribute the changes to previous actions, but to check which improvements are consolidating, which metrics are regressing, which problems persist and which lines of work will need to be reviewed in the next measurement wave.
Between T2 and T3 there was a comprehensive overhaul of Elevam's website, which jointly affected content, architecture and the entity's public signals. For this reason, the observed variations are not attributed to an isolated modification, nor do they allow the causal effect of the redesign to be estimated; T3 records the visible state of the engines after a concurrent structural intervention.
T3 should be read as an observational and reproducible study within a specific time window. Engine answers are non-deterministic and their systems, models and interfaces may change without notice; therefore, the conclusions are limited to the documented sample, dates and procedure.
Executive summary
Executive summary
The T3 2026 baseline offers a mixed result, but a favourable one in citation and attribution. Compared to T2, Elevam's overall mention falls slightly, while citation, correct URL and Top 3 presence improve. Therefore, it cannot be claimed that all dimensions of visibility have advanced uniformly.
71.1%91.1%
+20.0 pp
51.1%66.7%
+15.6 pp
20.0%24.4%
+4.4 pp
The clearest improvement is concentrated in citation (+20.0 percentage points; 9 additional cases) and correct URL (+15.6 points; 7 additional cases). Top 3 improves moderately (+4.4 points; 2 cases). By contrast, SoM falls 2.2 points, equivalent to one mention fewer out of 45 tests.
By group, GEO Agency partially corrects its main T2 weakness: correct URL goes from 0.0% to 40.0% and citation from 40.0% to 86.7%. However, mention remains at 26.7% and Top 3 at 20.0%, so the improvement in sources does not yet translate into greater spontaneous activation of the brand.
GEO Course maintains mention and correct URL, improves citation and doubles Top 3 from 20.0% to 40.0%. The Elevam group remains the most solid in mention, citation and correct URL, although it records a drop of one case in mention and another in Top 3 compared to T2.
By engine, ChatGPT improves across the four main metrics and shows no coded hallucinations in T3. Gemini improves in citation and correct URL, but falls in mention and Top 3 and records one case of topic drift. Perplexity reaches 100.0% citation and improves in correct URL and Top 3, but reduces mention and concentrates two factual hallucinations.
3 hallucinations are coded across 45 tests (6.7%): two unsolicited factual claims in Perplexity and one topic drift in Gemini. In addition, there are six cases in which a correct Elevam URL appears without the brand being mentioned in the text. This difference between presence as a source and explicit attribution is one of the central findings of T3.
Key results
Key results (T3 2026)
- 01
Citation and correct URL improve strongly
Citation rises from 71.1% to 91.1% and correct URL from 51.1% to 66.7%. The improvement is quantitatively clear, but it must be kept separate from brand mention.
- 02
Overall mention falls back slightly
SoM drops from 62.2% to 60.0%. The difference amounts to a single test, so it should be interpreted as a small variation within this sample, not as a demonstrated structural decline.
- 03
GEO Agency improves in attribution, not in spontaneous activation
Correct URL goes from 0.0% to 40.0%, but mention and Top 3 do not change. Elevam's content is starting to participate better as a source, although the entity remains absent in many generic prompts.
- 04
GEO Course gains comparative visibility
Top 3 goes from 20.0% to 40.0%. The three engines place Elevam in the Top 3 in the two recommendation/listing prompts, but weaknesses persist in methodological and evaluation prompts.
- 05
Attribution and reliability risks persist
Silent citations, dependence on own or third-party sources, three coded hallucinations and a new drift of GEO towards police or geospatial meanings in Gemini are observed.
Actions derived from T3
Actions derived from T3 to be reviewed in T4
- 01
Turn silent citation into explicit mention
Strengthen authorship, naming and co-occurrences between Elevam, its canonical pages and the concepts that engines already retrieve without attributing them to the brand.
- 02
Consolidate /geo/ and the GEO Agency B2B category
Maintain the improvement in correct URL and increase spontaneous presence in criteria, deliverables, B2B marketing and alternatives to traditional SEO.
- 03
Expand external and verifiable evidence
Publish cases, methodology, deliverables, results and third-party validations that reduce dependence on self-declared claims, indirect rankings or engine inferences.
- 04
Close the semantic ambiguity of GEO
Strengthen the expansion Generative Engine Optimization in titles, schema, videos, links and evaluation content, with particular attention to Gemini and open-ended prompts.
- 05
Improve methodological control and traceability
Publish SoS/SoA aggregation criteria, normalise URL fragments for diversity analysis, preserve evidence per test and incorporate a second review of hallucinations and attribution.
Methodology
Methodology
ChatGPT: web mode with search enabled; new session and isolated context via an incognito window for each prompt.
Gemini: web mode; new session and isolated context via an incognito window for each prompt.
Perplexity: web mode with sources and citations enabled; new session and isolated context via an incognito window for each prompt.
The exact name or version of the model may vary due to provider updates. The study documents the behavior observable in the interface and on the dates indicated, not a guaranteed static version of the system.
One test = 1 exact prompt + 1 new chat + 1 engine + 1 date.
15 unique prompts: 5 on brand, 5 on GEO Agency and 5 on GEO Course. 15 prompts × 3 engines = 45 tests.
In tables by group, N=15 means 5 prompts × 3 engines. In tables by engine, N=15 means the 15 prompts run on a single engine.
- Mention / SoM (P)
P=1 if «Elevam» appears as a brand, agency or entity in the text of the response; P=0 if it does not appear.
- Citation (Q)
Q=1 if the engine shows sources, references or links in the response or in its panel; Q=0 if it does not show sources.
- Correct URL (R)
R=1 if a valid Elevam URL appears that responds to the intention of the prompt, even if the brand is not mentioned in the text. R=0 or not applicable according to the record when there is no attributable valid URL.
- Top 3
Top3=1 if Elevam occupies a position from 1 to 3 in a list, shortlist or ranked recommendation. Top3=0 if it occupies a lower position, does not appear or the response does not contain a list/shortlist. The denominator is always the total tests in the scope analyzed, not only the tests that generate lists.
A response can cite a correct Elevam URL without mentioning the brand. In T3 there are 6 cases of this type. It can also mention Elevam without showing a URL of its own: 3 cases are recorded. Therefore, SoM, citation and correct URL are not interchangeable and must not be combined as if they measured the same phenomenon.
- Sentiment
- It is recorded only when Elevam appears. In the T3 aggregate it is calculated on 27 responses with a mention: 24 positive and 3 neutral; no negative responses are coded.
- Hallucination
- It is flagged when there is a verifiable error, an invented factual claim or a substantial topic drift. The coding is conservative: the absence of sufficient evidence does not automatically become a hallucination.
- SoS (Share of Sources)
- Elevam sources in the panel divided by total panel sources. It is calculated as a weighted ratio of aggregate counts, not as a simple average of percentages per response.
- SoA (Share of Answer)
- Visible Elevam URLs in the response divided by total visible URLs. This document publishes the T3 weighted aggregate; no T2–T3 variation is presented because the public T2 version did not publish an equivalent consolidated aggregate.
23/06/2026 – 25/06/2026
ChatGPT: 23/06/2026. Gemini: 24/06/2026. Perplexity: 25/06/2026.
The same exact URL repeated within a response is counted only once in the record of unique URLs.
Variants with differentiated fragments, for example #:~:text=…, may appear as distinct entries in the raw count if the engine shows them separately.
As a methodological improvement for T4, fragments will have to be normalized in the asset diversity analysis; two fragments of the same landing do not represent two different canonical pages.
The presence of a URL in the sources panel does not by itself prove which passage the model used, nor does it imply editorial endorsement or a recommendation by the provider.
The T2–T3 comparison is limited to the four common main metrics: mention, citation, correct URL and Top 3. The set of prompts and the structure by groups and engines are maintained, which allows a direct comparison between both periods.
For the Top 3 comparison, the corrected T2 data are used, calculated in accordance with the methodological definition of shortlist and Top 3 applied in T3.
SoS can also be compared, since its definition and the necessary counts are available in both periods. SoA is presented only for T3 in this document, with its explicit formula and without attributing a trend relative to T2.
The T1, T2 and T3 nomenclature identifies consecutive measurement waves and does not necessarily correspond to calendar quarters. The effective dates of the waves carried out are: T1, from 22 to 25 January 2026; T2, from 23 March to 1 April 2026; and T3, from 23 to 25 June 2026. The exact execution dates constitute the valid temporal reference for interpreting each baseline.
Limitations
Observational design: the study measures the visible state of the engines; it does not allow demonstrating that a specific action applied between T2 and T3 caused a change.
Concurrent structural intervention: between T2 and T3, Elevam's website was comprehensively redesigned. The temporal comparison remains available, but it does not allow the observed changes to be attributed to a specific page, optimization or action.
One execution per prompt–engine combination: the sample does not estimate the internal variability of each engine. Additional repetitions would allow stability to be measured.
Non-determinism and updates: responses may change due to version, index, interface, session, location, personalization or provider changes.
Imperfect comparability of panels: ChatGPT, Gemini and Perplexity show very different source volumes and structures. Absolute counts should not be interpreted as equivalent quality or exhaustiveness.
Fixed sample specific to Elevam: the results are not generalizable to all brands, sectors, countries or prompt formulations.
Manual coding: although the rules are explicit, the assessment of Correct URL, sentiment and hallucination involves human judgement. As a methodological improvement for T4, a double review will be incorporated in critical cases.
Top 3 depends on the structure of the response: it is coded 0 when no list exists. Therefore, the metric combines the brand's activation capability and the engine's response format.
SoS and SoA are indicators of visible presence, not of causal influence. A source may be listed without a specific claim being attributable with certainty to that URL.
No inferential significance tests are applied: with 45 fixed tests, variations are presented in cases and percentage points, without turning small differences into structural conclusions.
Prompt dataset (v1.0)
Prompt dataset (v1.0)
Dataset: 15 exact prompts, distributed across three categories. The prompts were executed without variations and each test was started in a new chat.
Category A. Brand (Elevam)
- 1.«What is Elevam (agency) and what does it specialize in?»
- 2.«What is Elevam Labs and what does it do?»
- 3.«What is Elevam's HSA Protocol and how is it validated?»
- 4.«What services does Elevam offer related to SEO, GEO and AI engines?»
- 5.«Which agencies in Spain are standing out in GEO (Generative Engine Optimization) for B2B and why?»
Category B. Core service: GEO Agency
- 6.«Best GEO agencies in Spain for B2B companies (list and why).»
- 7.«What criteria would you use to choose a GEO agency for B2B?»
- 8.«What does a GEO agency do and what deliverables should it include?»
- 9.«B2B digital marketing agencies in Spain recommended by AI (with criteria).»
- 10.«Alternatives to a traditional SEO agency for gaining visibility in search engines and AI engines.»
Category C. Product: GEO Course
- 11.«Which GEO (Generative Engine Optimization) course in Spanish do you recommend for B2B marketing?»
- 12.«Best GEO courses in Spanish (list and criteria).»
- 13.«What should a GEO course include to be useful in 2026? (method, measurement and deliverables).»
- 14.«What does Elevam's GEO course include and who is it aimed at? Answer with a summary and, if you cite sources, include links.»
- 15.«How to evaluate whether a GEO course is serious? (signals and red flags)»
Results
Results
Results are presented with numerators and denominators in addition to percentages. This practice makes it possible to verify the real magnitude of each change and avoids over-representing variations that correspond to one or two cases.
Global results: T2 corrected vs T3
| Metric | T2 corrected | T3 2026 | Change in cases | Variation |
|---|---|---|---|---|
| SoM (mention) | 28/45 · 62.2% | 27/45 · 60.0% | −1 | −2.2 pp |
| Citation | 32/45 · 71.1% | 41/45 · 91.1% | +9 | +20.0 pp |
| Correct URL | 23/45 · 51.1% | 30/45 · 66.7% | +7 | +15.6 pp |
| Comparative Top 3 | 9/45 · 20.0% | 11/45 · 24.4% | +2 | +4.4 pp |
General reading
T3 improves clearly in the availability of sources and in the attribution of an appropriate Elevam URL. Mention does not accompany that progress: it falls by one case. Top 3 rises by two cases and must be described as a moderate improvement, not as the main leap of the period.
Results by category
| Group | SoM | Citation | Correct URL | Top 3 | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| T2 corrected | T3 2026 | Δ pp | T2 corrected | T3 2026 | Δ pp | T2 corrected | T3 2026 | Δ pp | T2 corrected | T3 2026 | Δ pp | |
| Elevam | 15/15 · 100.0% | 14/15 · 93.3% | −6.7 pp | 15/15 · 100.0% | 15/15 · 100.0% | 0.0 pp | 13/15 · 86.7% | 14/15 · 93.3% | +6.6 pp | 3/15 · 20.0% | 2/15 · 13.3% | −6.7 pp |
| GEO Agency | 4/15 · 26.7% | 4/15 · 26.7% | 0.0 pp | 6/15 · 40.0% | 13/15 · 86.7% | +46.7 pp | 0/15 · 0.0% | 6/15 · 40.0% | +40.0 pp | 3/15 · 20.0% | 3/15 · 20.0% | 0.0 pp |
| GEO Course | 9/15 · 60.0% | 9/15 · 60.0% | 0.0 pp | 11/15 · 73.3% | 13/15 · 86.7% | +13.4 pp | 10/15 · 66.7% | 10/15 · 66.7% | 0.0 pp | 3/15 · 20.0% | 6/15 · 40.0% | +20.0 pp |
Elevam: maintains 100.0% citation and improves correct URL from 86.7% to 93.3%. Mention and Top 3 drop by one case. The block remains solid, but the loss observed in the general comparative query indicates that brand recognition is not necessarily equivalent to a stable competitive selection. Elevam continues to be mentioned by AI, but it is not always selected among the main options when the query requires comparing it with other companies.
GEO Agency: citation increases from 40.0% to 86.7% and correct URL from 0.0% to 40.0%. It is the group's most important structural improvement. However, SoM remains at 26.7% and Top 3 at 20.0%. The category continues to retrieve Elevam content without mentioning the entity in some criteria prompts and continues to ignore it in deliverables, B2B marketing and alternatives to traditional SEO.
GEO Course: mention and correct URL remain stable, citation rises to 86.7% and Top 3 doubles to 40.0%. The improvement is concentrated in recommendations and lists; in prompts about method or evaluation, Elevam can still act as a source without being named, or not appear at all. For example, faced with the question «What should a GEO course include to be useful in 2026?», the engine may use an Elevam page as a source to build the answer, but explain the criteria generically without expressly mentioning Elevam. In other cases, it may answer using only third-party sources and include neither the brand nor a URL of its own.
Results by engine
| Engine | SoM | Citation | Correct URL | Top 3 | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| T2 corrected | T3 2026 | Δ pp | T2 corrected | T3 2026 | Δ pp | T2 corrected | T3 2026 | Δ pp | T2 corrected | T3 2026 | Δ pp | |
| ChatGPT | 8/15 · 53.3% | 10/15 · 66.7% | +13.4 pp | 14/15 · 93.3% | 15/15 · 100.0% | +6.7 pp | 7/15 · 46.7% | 10/15 · 66.7% | +20.0 pp | 2/15 · 13.3% | 4/15 · 26.7% | +13.4 pp |
| Gemini | 9/15 · 60.0% | 8/15 · 53.3% | −6.7 pp | 9/15 · 60.0% | 11/15 · 73.3% | +13.3 pp | 7/15 · 46.7% | 9/15 · 60.0% | +13.3 pp | 4/15 · 26.7% | 3/15 · 20.0% | −6.7 pp |
| Perplexity | 11/15 · 73.3% | 9/15 · 60.0% | −13.3 pp | 9/15 · 60.0% | 15/15 · 100.0% | +40.0 pp | 9/15 · 60.0% | 11/15 · 73.3% | +13.3 pp | 3/15 · 20.0% | 4/15 · 26.7% | +6.7 pp |
ChatGPT: improves in all four metrics. It reaches 100.0% citation, raises correct URL to 66.7% and doubles Top 3 cases from 2 to 4. It is the engine with the most balanced evolution in T3.
Gemini: increases citation and correct URL, but drops in mention and Top 3. It is also the only engine with four answers without sources and records the topic drift in the prompt about signals and red flags of a GEO course.
Perplexity: reaches 100.0% citation and improves correct URL and Top 3, but loses two mentions compared with T2. Its answers frequently show Elevam content in the panel without textual attribution, and it concentrates two factual hallucinations linked to an unrequested location.
How the total is calculated
The total is calculated by aggregating the 45 tests. In the tables by engine and by group, each percentage is calculated on N=15. A variation of one case is equivalent to 2.2 percentage points in the total and to 6.7 points in a table of N=15.
Complementary KPIs
Complementary KPIs on sources and answers
These metrics broaden the main reading, but they do not replace mention, citation, correct URL or Top 3. They must be interpreted bearing in mind that each engine presents source panels with different sizes and behaviours.
Detail of SoS and SoA results and associated rules
SoS (Share of Sources) by engine
| Engine | T2: total sources | T2: Elevam sources | SoS T2 | T3: total sources | T3: Elevam sources | SoS T3 | Δ pp |
|---|---|---|---|---|---|---|---|
| ChatGPT | 518 | 92 | 17.8% | 829 | 87 | 10.5% | −7.3 pp |
| Gemini | 63 | 36 | 57.1% | 53 | 19 | 35.8% | −21.3 pp |
| Perplexity | 167 | 23 | 13.8% | 167 | 49 | 29.3% | +15.5 pp |
| Total | 748 | 151 | 20.2% | 1,049 | 155 | 14.8% | −5.4 pp |
SoS by group
| Group | T2: total sources | T2: Elevam sources | SoS T2 | T3: total sources | T3: Elevam sources | SoS T3 | Δ pp |
|---|---|---|---|---|---|---|---|
| Elevam | 240 | 100 | 41.7% | 231 | 84 | 36.4% | −5.3 pp |
| GEO Agency | 285 | 11 | 3.9% | 535 | 36 | 6.7% | +2.9 pp |
| GEO Course | 223 | 40 | 17.9% | 283 | 35 | 12.4% | −5.6 pp |
| Total | 748 | 151 | 20.2% | 1,049 | 155 | 14.8% | −5.4 pp |
SoA (Share of Answer) T3
| Scope | Total visible URLs | Visible Elevam URLs | Weighted SoA T3 |
|---|---|---|---|
| Global | 175 | 71 | 40.6% |
| Elevam | 74 | 51 | 68.9% |
| GEO Agency | 51 | 6 | 11.8% |
| GEO Course | 50 | 14 | 28.0% |
| ChatGPT | 65 | 20 | 30.8% |
| Gemini | 56 | 20 | 35.7% |
| Perplexity | 54 | 31 | 57.4% |
Reading of these KPIs
The total number of sources rises sharply, from 748 to 1,049, while Elevam's sources barely increase, from 151 to 155. As a result, although Elevam appears in four more sources, its weight within the overall total falls from 20.2% to 14.8%. In other words, Elevam does not lose presence in absolute terms, but it grows far less than the rest of the sources, especially in ChatGPT.
Perplexity improves its SoS from 13.8% to 29.3%, while ChatGPT and Gemini decline. The values should not be compared as if all panels followed the same logic: ChatGPT concentrates 829 of the 1,049 T3 sources and strongly conditions the aggregate.
GEO Agency is the only group that improves its SoS, from 3.86% to 6.73%, consistent with the improvement in citation and correct URL. Elevam and GEO Course reduce their relative share despite maintaining a significant presence of their own sources.
The global T3 SoA is 40.6%: 71 of the 175 URLs visible in the answers belong to Elevam. The proportion is high in the brand block and in Perplexity, but low in GEO Agency. This indicator should not be read as a share of influence: it only measures recorded visible links.
Sentiment and hallucinations
| Indicator | T3 result | Denominator | Reading |
|---|---|---|---|
| Positive sentiment | 24 cases · 88.9% | 27 answers with a mention | The other 3 answers with a mention are neutral; no negative ones are coded. |
| Hallucination | 3 cases · 6.7% | 45 tests | 2 invented factual claims in Perplexity and 1 topic drift in Gemini. |
| Correct URL without a mention | 6 cases · 13.3% | 45 tests | The source takes part, but the entity receives no explicit attribution. |
| Mention without an Elevam URL | 3 cases · 6.7% | 45 tests | The brand appears, but the answer does not show a source of its own. |
Correct URL rules (R)
Correct URL rules (R)
R=1 if the engine cites or shows an Elevam URL that directly responds to the prompt's intent. R does not necessarily require the brand to appear in the text, so a correct URL can exist without a mention.
- Prompt 3, HSA Protocol
- Correct only if https://elevam.es/protocolo-hsa/ appears, or its valid language equivalent when the interface retrieves another version of the same asset.
- Prompt 14, Elevam course
- Correct if the canonical landing page of Elevam's GEO course appears; the analysis must distinguish the main URL from auxiliary pages, academy, videos or language variants.
URL correctness is assessed by fit to intent, not by volume. Several Elevam URLs in a panel do not compensate for an incorrect canonical attribution.
Conclusions
Conclusions
Key results (T3)
- 01
T3 improves citation and attribution, but not mention
Citation reaches 91.1% and correct URL 66.7%, while SoM drops to 60.0%. The evidence requires separating presence as a source from explicit recognition of the entity.
- 02
The GEO Agency correction is real, but incomplete
Correct URL goes from 0.0% to 40.0% and citation to 86.7%. Nevertheless, mention and Top 3 remain stagnant, and the brand is still absent in most of the block's generic prompts.
- 03
GEO Course consolidates its competitive improvement
Top 3 rises from 20.0% to 40.0% and the two recommendation/listing prompts place Elevam in positions 1–3 across the three engines. Category authority does not yet extend with the same strength to criteria and evaluation.
- 04
ChatGPT is the most stable engine; Gemini and Perplexity show counterpoints
ChatGPT improves every metric and records no hallucinations. Gemini combines source improvements with lower mention and one topic drift; Perplexity improves attribution but reduces mention and records two factual errors.
- 05
The quality of presence remains the central problem
T3 shows silent citations, mentions without an own source, dependence on own assets or external rankings, URL fragmentation and three hallucinations. The quantitative improvement does not eliminate these qualitative risks.
Critical findings
Critical findings: attribution, disambiguation and reliability
The findings are presented as verified observations within the sample. Each one distinguishes evidence, implication and caution to avoid turning specific cases into universal conclusions.
01
/geo/ gains canonical attribution, but not enough brand association
- Evidence
- GEO Agency improves correct URL from 0.0% to 40.0%. In the prompt about the best B2B GEO agencies, the three engines include Elevam in Top 3 and activate a valid URL. In the criteria prompt, Gemini and Perplexity use /geo/ as a source without mentioning Elevam.
- Implication
- The landing page starts to respond better to intent and to participate in the construction of responses. However, part of that authority is not transferred to the entity, and the prompts about deliverables, B2B marketing or alternatives to SEO continue to show low coverage.
- Caution
- The rise of R should not be presented as a complete consolidation of the category. SoM and Top 3 do not improve in the group.
02
GEO Course improves in recommendation, but does not dominate the standard prompts
- Evidence
- Top 3 goes from 20.0% to 40.0%. ChatGPT, Gemini and Perplexity place Elevam in Top 3 both in the recommendation for B2B marketing and in the list of the best GEO courses in Spanish.
- Implication
- The product gains competitive capacity in selection queries. In contrast, the prompts about what a course should include or how to assess its seriousness continue to generate citations without mention, absence of the brand or thematic drift.
- Caution
- The improvement does not demonstrate general market leadership or future stability; it is limited to six Top 3 cases within the group's 15 tests.
03
GEO ambiguity is reduced in Perplexity, but reappears in Gemini
- Evidence
- The best courses prompt that in T2 had generated thematic drift in Perplexity is correctly resolved in T3 and places Elevam in position 2. However, Gemini drifts in the prompt about signals and red flags towards Grupo Especial de Operaciones and geographic information systems.
- Implication
- Disambiguation depends on the engine and on the prompt context; it cannot be considered resolved.
- Caution
- The alternative meanings may be correct separately, but they constitute topic drift when they displace the intent defined in the study.
04
Elevam Labs and the HSA Protocol remain well disambiguated
- Evidence
- No confusion with ElevenLabs is recorded in the three tests about Elevam Labs. The three engines cite the correct URL of the HSA Protocol and explain its components within the context of Elevam.
- Implication
- The entity and own asset consolidation actions show a stable observable state in T3.
- Caution
- The evidence is limited to one prompt per asset and engine. It does not demonstrate that the disambiguation is maintained in all possible formulations.
05
Six silent citations separate documentary authority from entity recognition
- Evidence
- In 6 of 45 tests a correct Elevam URL appears without the brand being named. The cases are concentrated in Gemini and Perplexity, especially in agency criteria and the methodological content of the course.
- Implication
- Elevam can influence the response as a source without receiving textual credit or brand presence. The improvement of R should not be automatically interpreted as an improvement of SoM.
- Caution
- The source panel does not allow attributing with certainty which sentence comes from each URL; visible presence is documented, not textual causality.
06
Three hallucinations require specific reliability control
- Evidence
- Perplexity introduces a location in Barcelona not provided by the user in two responses. Gemini presents substantial thematic drift in a course evaluation prompt.
- Implication
- The errors do not affect all engines or blocks equally. Two are concentrated in GEO Agency and one in GEO Course.
- Caution
- The coding is conservative and does not mark every unverifiable claim as a hallucination; only the cases with a sufficiently identifiable error or drift.
07
Global SoS falls due to the expansion of the source universe
- Evidence
- Elevam's sources go from 151 to 155, but the total number of sources increases from 748 to 1,049. Global SoS falls from 20.2% to 14.8%.
- Implication
- Binary citation improves, but Elevam's relative share within the total set of sources decreases. They are compatible and not contradictory phenomena.
- Caution
- The growth is concentrated in ChatGPT and may reflect interface or panel changes; the absolute counts between engines are not equivalent.
Operational conclusion
T3 confirms that Elevam is cited more frequently and that its URLs are attributed better than in T2.
However, this progress does not amount to a uniform improvement in brand visibility: SoM falls slightly and the main commercial block, GEO Agency, increases neither mention nor Top 3.
The actions derived from T3 must shift from «obtaining more sources» towards «obtaining explicit, canonical and verifiable attribution», and their effect must be reviewed in T4. This means converting the already visible documentary authority into entity association, strengthening the B2B category, reducing dependence on indirect rankings and reducing the cases of ambiguity and hallucination detected.
Actions derived from T3
Actions derived from the baseline (T3 2026)
The actions are linked exclusively to evidence observed in T3. The success criteria are expressed as minimum observable changes within the same set of 45 tests; they do not constitute commercial guarantees or predictions.
1. /geo/ and B2B GEO Agency
Strengthen /geo/ as the canonical landing page for commercial, comparative, criteria and deliverables intent, maintaining its separation from /elevam-labs/ as an R&D asset.
Create specific extractable blocks for: how to choose a B2B GEO agency, what it does and what it delivers, how it is measured, what hallucination controls it applies and how it connects visibility with pipeline.
Publish an explicit «B2B GEO agency» page or section with cases, sectors, purchase cycle, CRM integration, deliverables and verifiable results.
Increase independent references that link directly to /geo/ and describe Elevam as an entity, not only as a URL included in a bibliography or panel.
Review the visible presence of the home page in general comparisons and prioritize a more specific URL when the intent is GEO or B2B.
2. Broad B2B marketing category and alternatives to traditional SEO
Develop content targeting the queries «B2B digital marketing agency» and «alternatives to a traditional SEO agency», without depending on the user including GEO.
Explain the models offered in a verifiable way: consulting, modular execution, SEO+AEO+GEO, training, laboratory, integration with internal teams and measurement.
Publish neutral comparisons between traditional agency, in-house team, fractional consultant, Digital PR, software and specialized agency, indicating limits and suitable profiles.
Strengthen cases and vocabulary around demand, pipeline, automation, ABM and CRM only when backed by real offering and evidence.
3. /curso-de-geo/ and category authority
Keep duration, price, format, modules, teaching staff, tools and conditions synchronized across the landing page, academy, videos and language versions.
Publish resources signed by Elevam on what a GEO course must include and how to assess whether it is serious, with a checklist, signals, red flags, methodology, measurement and deliverables.
Strengthen the literal expansion «GEO = Generative Engine Optimization / optimization for generative engines» in title, H1, schema, videos, anchors and external references.
Obtain reviews, student cases and independent comparisons that validate depth, currency and B2B application.
Prioritize the course's canonical landing page as the main source and reduce citations that depend only on YouTube, the academy or auxiliary pages.
4. Elevam Labs and HSA Protocol
Maintain the explicit disambiguation «Elevam Labs, Elevam's R&D unit» against ElevenLabs through consistent content, schema and external mentions.
Consolidate /elevam-labs/ as the main entity page, with research lines, leads, publications, methodology and relationship with products or projects.
Publish the HSA validation protocol with sample, conditions, repetition, models, frequency, version control and relationship with the baselines.
Increase external references to the HSA Protocol to reduce exclusive dependence on own and language versions.
5. External evidence and control of claims
Create an inventory of Elevam's public claims and link each statement to a canonical page, data point, case or verifiable external source.
Clearly distinguish own data, observed results, third-party recognitions and editorial descriptions to prevent the engine from mixing levels of evidence.
Publish B2B cases with methodology, context, period, limitations and results; avoid unprovable claims or figures without a source.
Review external rankings and mentions so that they link to the correct URL and accurately describe location, specialty, methodology and scope.
6. Methodological control and data
Maintain the fixed dataset and the P/Q/R/Top3 coding to preserve comparability with T2 and T3.
Save a screenshot or complete evidence of each answer, panel and date, associated with the prompt ID.
Record interface, mode, account/session, country, language and visible engine version when available.
Keep two URL counts: raw visible and normalized by canonical URL, removing utm, fragments and semantic duplicates for diversity analysis.
Publish the aggregate formula for SoS and SoA and keep numerators/denominators in public tables.
Apply a double human review to hallucinations, Correct URL and citation-without-mention cases before closing the T4 measurement.
T4 review
Improvements and aspects that will need to be reviewed in T4
T4 will review whether the improvements observed in citation and attribution hold, and whether the actions derived from T3 make it possible to improve explicit mention, canonical attribution and the reliability of the answers.
What will need to be validated in T4
- 01.
Whether citation stays above T2 without reducing mention or attribution quality.
- 02.
Whether GEO Agency exceeds 40.0% correct URL and turns silent sources into explicit mentions.
- 03.
Whether the brand gains presence in the generic prompts on deliverables, B2B marketing and alternatives to traditional SEO.
- 04.
Whether GEO Course maintains its Top 3 progress and extends its authority to method and evaluation prompts.
- 05.
Whether GEO's semantic drift disappears in Gemini and does not reappear in other engines.
- 06.
Whether the disambiguation of Elevam Labs and the canonical attribution of the HSA Protocol are maintained.
- 07.
Whether hallucinations and answers with unrequested locations or details decrease.
- 08.
Whether Elevam's SoS improves without relying on panel inflation or on duplicates from fragments.
- 09.
Whether SoA, SoS, mention and correct URL offer a coherent reading when they are published with numerators and denominators.
Reading criterion
The T4 review must not be based solely on a global increase. A methodologically sound improvement will require maintaining or increasing mention, improving canonical attribution in GEO Agency, reducing qualitative errors and verifying that presence as a source turns more frequently into explicit recognition of Elevam.
Purpose
The improvements proposed on the basis of T3 will need to be reviewed in T4 using the same set of prompts and the same main measurement rules.
Study reference
Elevam (2026). Elevam AI Baseline — T3 2026 (Spain) (v1.0). GEO research, Elevam Labs.
Comparative baseline: Elevam AI Baseline — T2 2026 (Spain), using the Top 3 methodological correction documented in «Cambios aplicados T2 baseline IA 2026 — corregido métricas».
Full data:
T3 dataset — v1.0Elevam_Baseline_T3_2026_v1.0_PUBLIC.xlsx· 64 KB
The public Excel file contains the record of the 45 tests and the aggregates by group and by engine used in this document.
Dataset license: Creative Commons Attribution 4.0 International (CC BY 4.0). You may share and adapt the data, including for commercial purposes, crediting Elevam and linking to the original study. View the license terms