Study 3 — Validation Interval 5: Proof-Based UK GEO Specialist Comparison Query

On 19 September 2026, NeuralAdX Ltd ran the sixth test in Study 3 across Google AI Mode, Perplexity, Microsoft Copilot and ChatGPT. Google AI Mode produced a strong NeuralAdX result with a first-sentence brand mention, multiple NeuralAdX citations and three NeuralAdX URLs visible in the source panel. Perplexity included NeuralAdX as a GEO specialist with publicly testable proof, although another company received greater prominence. Microsoft Copilot also prioritised another company. ChatGPT produced a strong NeuralAdX result with multiple NeuralAdX citations, repeated brand mentions and substantial use of NeuralAdX evidence in the generated answer.

Hello and welcome back again.

As always, it is Paul Rowe, Founder, Chief Generative Engine Optimisation Officer and CEO at NeuralAdX Ltd.

We are now running the third prompt in our live AI retrieval testing across the usual four platforms: Google AI Mode, Perplexity, Microsoft Copilot and ChatGPT.

This query was first tested in January 2026, and this is the sixth time we have tested it.

The purpose is to observe citation stability and AI visibility over time, including brand mentions, citation absorption and, where possible, citation fidelity.

Citation fidelity will need to be assessed separately by comparing the generated answer directly with the original source material.

Today is 19 September 2026.

So this test takes place approximately nine months after the original test.

Let us see how NeuralAdX performs in this longitudinal Generative Engine Optimisation testing.

We will begin with Google AI Mode.

That is an excellent result.

Straight away, within the first sentence of the AI summary, NeuralAdX Ltd receives a brand mention.

It also identifies me directly as specialist Paul Rowe and provides a link associated with my name.

The response then expands with detailed information about the testing and evidence published by NeuralAdX.

We have two NeuralAdX domain citations visible here.

Looking at the source panel on the right-hand side, three NeuralAdX URLs are being used as sources.

For this answer, Google AI Mode is therefore relying heavily on NeuralAdX content to support its response.

Moving further down, NeuralAdX Ltd is mentioned again.

There are then three additional visible source references associated with NeuralAdX content.

Some of these may correspond with the same URLs displayed in the source panel.

Overall, this is a very strong response approximately nine months after the original test.

There is a reasonable amount of citation absorption, with multiple NeuralAdX citations and repeated brand mentions throughout the answer.

Now let us look at Perplexity using the same query.

Perplexity uses a different retrieval and answer generation system from Google AI Mode, so the source selection can vary.

This time, the result is different.

Another company is prioritised more prominently in the answer.

That is an important reminder that AI engines can choose different sources for the same query.

However, NeuralAdX is still included.

Looking further down, the answer includes a section referring to other UK specialists with strong testable proof.

NeuralAdX Ltd is identified as a GEO specialist with publicly testable proof.

The answer references our live AI retrieval testing and supporting evidence.

Interestingly, the citation supporting this mention is from a third-party source rather than directly from NeuralAdX.

That may indicate that Perplexity is giving particular weight to independent third-party evidence for this type of comparative query.

That is useful information in itself.

For this test, NeuralAdX has not achieved the most prominent position in Perplexity, but the brand is still included and associated with publicly testable GEO evidence.

So this is a weaker result than Google AI Mode, but still a meaningful surfacing.

Now we move to Microsoft Copilot.

As I have mentioned in other testing videos, we have experienced ongoing issues with Bing Webmaster Tools and the way our business information is recognised within the Microsoft ecosystem.

That may influence our visibility in Microsoft Copilot, although this individual test cannot establish the precise cause of any result.

In this answer, another GEO company is surfaced prominently.

The answer references claims made on that company’s website regarding the number of businesses or assessments completed.

This demonstrates an important feature of AI search.

AI systems may incorporate publicly available company claims into generated answers, particularly when those claims are clearly stated and easily retrievable.

However, the presence of a claim in an AI-generated answer does not independently verify the underlying claim.

That distinction is important when assessing AI-generated recommendations.

For this Microsoft Copilot test, NeuralAdX does not receive the leading position.

So the observed result here is significantly weaker than the Google AI Mode result.

Now let us move to ChatGPT using exactly the same query.

I can already see NeuralAdX appearing within the source selection.

The system is retrieving and analysing several sources before generating the answer.

Let us see whether NeuralAdX receives strong brand visibility and citation coverage.

The answer states that, based on public evidence it could verify as of 19 September 2026, NeuralAdX Ltd currently has the most comprehensive publicly documented proof stack among UK Generative Engine Optimisation specialists.

That is a very strong result for the specific query being tested.

The answer then provides substantial supporting detail.

It says that the distinction is the breadth and repeatability of the evidence.

It references NeuralAdX publishing multiple live proof videos, structured queries, screenshots, transcripts and platform-level citation evidence.

This is a strong example of citation absorption because a significant amount of information from NeuralAdX source material has been incorporated directly into the generated response.

There is also a NeuralAdX domain citation.

Then another NeuralAdX domain citation.

And another NeuralAdX domain citation.

The answer later introduces other entities before returning to NeuralAdX again.

There is another NeuralAdX brand mention and another domain citation.

Further down, NeuralAdX is mentioned again with another citation.

So this is a very strong ChatGPT performance.

The important point is not simply that NeuralAdX appears.

The AI answer is incorporating a substantial amount of supporting information from NeuralAdX content and using that evidence to explain its conclusion.

That is highly relevant when assessing citation absorption.

Citation fidelity is a separate measurement.

To establish citation fidelity properly, I would need to compare the wording and factual claims in the AI-generated response directly against the original NeuralAdX source pages.

Overall, this test demonstrates considerable variation across the four platforms.

Google AI Mode produced a very strong NeuralAdX result with multiple citations and extensive source usage.

Perplexity included NeuralAdX and recognised its publicly testable evidence, but another company received greater prominence.

Microsoft Copilot also prioritised another company in this test.

ChatGPT produced a very strong NeuralAdX result with extensive citation absorption, multiple domain citations and detailed reference to our published evidence.

That variation is exactly why longitudinal, multi-platform AI retrieval testing is useful.

Generative Engine Optimisation performance should not be assessed from a single prompt, a single AI platform or a single point in time.

Thank you very much for watching.

As always, I look forward to seeing you in a future video.

Thank you and take care.

Bye bye.