Book an AI Intelligence audit
Standalone audit, or a monthly service from £899 a month plus VAT
What AI is saying about your organisation, AND what to do about it
Viva AI Intelligence is Viva’s reputation service for a world in which AI assistants are part of how people understand an organisation. We audit what ChatGPT, Gemini, Copilot and Claude say about you, grade the evidence behind it, and tell you which communications actions are worth taking. Reports are delivered under the Terrain name.
People now put questions to AI assistants that they once put to search engines or colleagues: researching an organisation, comparing suppliers, understanding a complex subject, deciding who to shortlist. Buyers of PR do it too. They paste a brief into an LLM and ask for five agencies to consider.
We ran the audit on ourselves in October and November 2026. The assistants named Viva on more agency shortlists than any rival in our sectors, described us consistently, and still missed two things we would want a buyer to know. This page was one of the changes that followed.
A GEO audit, also called an AI visibility or AI search audit, measures how often you appear and what is said. That is useful and it is only the start. Viva AI Intelligence goes on to check whether what is said is accurate, where it comes from, who else is named and why, and which gaps are worth closing. The judgement is the service.
Viva AI Intelligence (formerly the AI Awareness Baseline) combines structured AI search analysis with communications expertise to show how your organisation is represented, which evidence is shaping that picture, where misinformation or reputation gaps exist and what communications should do about them.
What a Viva AI Intelligence audit looks at
We analyse the themes, capabilities and reputation territories AI assistants associate with your organisation. A dashboard shows whether your name appears. We look at how you are being understood and described, and whether that matches the position you want to hold.
How it works.
We ask a series of questions to the LLMS over time to get a direction of travel. We use questions with both your brand name and unbranded questions. The latter can be particularly revealing.
If someone asks an AI assistant about the kind of problem you solve without mentioning your company name, do you naturally enter the conversation? And if you don’t, who does?
We examine claims for information that appears inaccurate, outdated, unsupported, inconsistent or potentially misleading. Where something looks wrong, we investigate the evidence behind it rather than simply flagging an awareness score.
AI answers do not exist in isolation. They draw on an information environment containing earned media, corporate websites, government information, reviews, research, social content, forums and other sources.
We identify the sources appearing in relevant answers and assess what they actually tell us about your reputation.
An external website is not automatically independent evidence. A client independently confirming a successful project, for example, tells us something very different from a company describing its own capabilities on its website. We can also find gaps and out-of-date information from your own website that might still be surfacing on LLMs.
In essence, our analysis looks at the quality, relevance and independence of the evidence supporting important reputation claims.
Sometimes the important finding is not that a competitor appears more often – It is why.
We look at where competitors appear to have stronger evidence, clearer associations, or greater ownership of strategically important subjects.
What you should do next.
Every finding comes with a decision. We tell you whether to protect something that is already working, clarify something the assistants are getting wrong, strengthen a gap where you should be named and are not, or leave it alone because it does not matter to your objectives.
The actions are things a communications team can deliver. Typically they include correcting outdated information, publishing results that exist but are not public, making existing evidence easier to find, giving an under-recognised area of expertise a page of its own, or working with a specific trade title or stakeholder group. For each priority we name the claim or gap, the content or channel that addresses it, who produces it, and the improvement to look for at the next check.
Then we re-test, so you can see whether the assistants have picked it up.
How a Viva AI Intelligence audit works.
AI answers vary. Ask the same question on the same platform a day later, or with slightly different wording, and the sources, the framing and the organisations recommended can all change. Viva’s method is built around repeated, consistent observation and retained evidence, so a single unusual answer is never mistaken for a pattern.
Viva’s methodology is therefore designed around consistent measurement, transparency, and evidence rather than false precision.
Our thinking is guided by the AMEC GEO Principles and the wider AMEC approach to communications measurement and evaluation.
AMEC recommends measuring AI-led discovery across three connected areas: the upstream reputation environment, search and content readiness and downstream AI outputs. It also recommends governed query sets, repeated observations, saved evidence, source-quality assessment and clear separation between visibility and genuine communications outcomes.
In practice, Viva:
- Agrees the stakeholder questions and reputation territories worth understanding.
- Builds a structured set of questions a real customer, journalist or buyer would ask, some naming you and some not.
- Tests those questions systematically across selected AI platforms.
- Repeats observations where appropriate so that one unusual answer is not mistaken for a trend.
- Retains relevant answers and source evidence.
- Analyses reputation, framing, accuracy, misinformation and evidence quality.
- Identifies meaningful gaps and opportunities.
- Turns the findings into a plain-English communications strategy.
- Re-tests after the actions are taken, so the client can see what changed.
It is a Viva methodology guided by AMEC’s GEO Principles and established communications measurement practice.
What you get.
A Viva AI Intelligence audit is built to answer five questions:
- What is AI saying about us?
- Where are we absent, misunderstood or weak?
- What evidence is shaping that picture?
- What is wrong, outdated or poorly supported?
- What should communications do next?
GEO measurement without the jargon
There are increasingly sophisticated tools available for tracking AI visibility. We use technology where it makes the work faster and more reliable. But the technology is not the strategy.
Our role is to turn the information into communications advice that people can understand and actioned.
Want to explore your AI reputation?
Viva AI Intelligence is available as a standalone audit or as a monthly service from £899 a month plus VAT, with quarterly re-tests. If you want to know what AI is saying about your organisation and what to do about it, talk to us.
Frequently Asked Questions (FAQ) About GEO and AI Search Audits:
GEO stands for Generative Engine Optimisation. A GEO audit looks at how a brand, organisation, product, issue or subject appears in AI-generated search and answer systems.
A good audit should go beyond simply measuring mentions. It can examine how an organisation is described, which sources shape the answers, whether information is accurate, and where communications opportunities or reputation risks exist Viva’s audit, delivered as Viva AI Intelligence, does all of those.
These terms are relatively new and you will see often see then used interchangeably. We use GEO audit, AI search audit and AI visibility audit where it makes sense because the market is still settling on its language.
Our own approach places particular emphasis on reputation, accuracy, evidence and communications action rather than visibility alone.
The appropriate platforms depend on the organisation, audience and project. Major generative search and answer systems can behave differently, which is why we do not assume one platform represents the whole AI environment.
We can, but we won’t. We can measure visibility – but we do not believe one number adequately represents an organisation’s AI reputation.
We’re not the only ones. AMEC‘s GEO Principles similarly caution against treating a single score, tool or query set as the whole picture. We therefore use visibility as one piece of evidence rather than the final answer.
Because generative AI answers vary. The same broad question can generate different answers at different times or on different platforms. Repeated observations help us identify persistent patterns rather than overreact to one unusual result.
No GEO audit can tell you exactly what every user will see in every possible AI interaction. You will get different scores from different tool providers too. What we can do is apply a governed, repeatable methodology to a defined set of stakeholder questions and use those observations to identify meaningful patterns.
Again, we look to AMEC here and it describes observed AI outputs as directional indicators that should be tested transparently across platforms, prompts and time.
That’s a question for both of us. We start with objectives, audiences, reputation priorities and genuine stakeholder information needs.
Questions can include both unbranded queries, which test whether an organisation emerges naturally, and branded queries, which examine what AI systems believe and say when the organisation is explicitly discussed.
Where possible, we also use search behaviour, customer questions and other demand signals to inform the query set.
No. Nobody can. That’s the honest truth. There is currently no public equivalent of Google Trends or Answer The Public containing reliable search volumes for every prompt submitted to every major AI platform. We don’t pretend that an estimated prompt set represents the entire population of AI questions.
AI needs to have the human at the helm but what an AI audit can help with is identify claims that appear inaccurate, outdated, inconsistent, unsupported or misleading. However, findings need to be verified.
We can also use the method to look at the sources associated with questionable claims. This would obviously be useful for communications teams who want to understand why a particular version of the story may be appearing.
AI systems draw information from many different sources. A company’s own website can be useful and authoritative for factual information, but independent evidence can play a different reputational role.
Credible media coverage, a customer statement, government record, expert source or industry reference may independently substantiate something the organisation says about itself.
AMEC‘s GEO Principles emphasise the importance of reliable, trustworthy and current sources rather than simply maximising the volume of material available online.
It’s not that simple. We can sometimes identify earned media among the sources associated with AI answers. That provides useful evidence that PR-generated material has entered the public information environment AI systems are using.
We would be cautious about claiming cause and effect without strong supporting evidence. We can show which sources the assistants cite and whether coverage appears among them, and we can find out with a re-test if new coverage shows whether the answers changed, but we can’t claim causation.
Search (SEO) still matters not least because many AI systems use web search or retrieval as part of producing answers. A lot of their information is drawn from your own website.
But GEO also involves reputation evidence beyond the organisation’s website, including earned media, reviews, public information and third-party commentary.
GEO adds another evidence source. AI-led discovery should ultimately be connected back to communications objectives, stakeholder understanding, reputation and organisational outcomes rather than treated as a standalone success metric.
There is no single answer. For many organisations, a baseline followed by periodic comparable reviews may be more useful than reacting to daily changes.
Organisations dealing with a live issue or misinformation risk may want more frequent observation.
Our general principle is simple: do not change a communications strategy merely because one AI measurement moves.
Yes. Viva has developed a structured methodology for GEO audits and AI reputation analysis and is applying it to selected organisations and communications challenges. We call it an AI Awareness Baseline tool.
If you want to explore whether it would be useful for your organisation, speak to us and we can discuss an appropriate scope.
It depends on the number of reputation areas, questions, platforms, markets and depth of analysis required. We scope projects according to the communications question that need answering, rather than selling a fixed one-size-fits-all package.
We start with communications rather than a dashboard. Our interest is not simply “how visible are you?” it’s “What is AI saying about you, what evidence is shaping that reputation, what looks wrong or weak, and what should communications do about it?” We use the technology to gather evidence, but our value is knowing what to do with it.
Well, while you can change the public evidence AI draws on, you cannot engineer an answer. But if you change what you have control of you re-test to see what moved.
No. We only use public information. We use our deep domain knowledge to shape our insights.
Terrain is the name of the report, it contains our findings and, critically, three actions for you to take.