Search as we know it is collapsing. Google traffic is down for the first time in its history – confirmation that people aren’t starting their journeys on the traditional web anymore. They’re asking ChatGPT, Perplexity, or Gemini. They’re getting a single, neat answer instead of ten blue links. And the majority of the time, they’re satisfied with the response they get.
The marketing industry is in the middle of a generational panic in response. As New York magazine put it: “For tens of thousands of SEO gurus, Google whisperers, and marketing professionals who work on and around search – a group of people accustomed to sudden changes at the whims of software giants – it’s been chaos.”
After the shock of change, a new mantra is emerging both inside and outside marketing departments – we must make sure our company appears when people search for us, and that what these programs say about our company is accurate. Fair enough – but how?
A more systematic approach is needed. Businesses need an audit.
The slow (and then fast) death of traditional SEO
SEO used to be a game of mechanics: load up on keywords, stack enough backlinks, and you could claw your way up the rankings. That world is gone. AI has broken the old rules of discoverability.
Even though AI is still described as an “emerging” technology, it is already a dominant force shaping buying decisions. One recent study found that 89% of B2B users now consider information surfaced by large language models (LLMs) when making search-related decisions. And 80% of users report being happy with AI-generated summaries without ever clicking through to a website.
Behaviour is shifting faster than the industry can adapt. Search behaviour through LLMs looks very different from the old search bar. Queries are longer and more specific: Perplexity queries average 11-12 words, compared to 3-4 on Google. We’re creating a whole new grammar for interacting with these bots. Development cycles of six to nine months per major product release can’t keep up.
And LLMs don’t simply match words to queries, filtered through backlinks. They digest meaning – making purposeful decisions about who to trust and who to ignore.
This makes the discovery problem both more subtle and more challenging. LLMs are highly sophisticated and increasingly personalised. Where “Googling yourself” was once a crude but reliable proxy for understanding your online footprint, that exercise no longer works. Asking a model “What is [Company X] known for?” might yield a different answer depending on your history, your phrasing, or your location.
Like Google search results, LLM outputs are fragmented. No single person sees the exact same thing. The result is a discoverability landscape that is harder to control, harder to measure, and, for now, easier to get wrong.
What is a GEO Audit?
With AI search ascendant, we need accurate measurement tools to understand its reach and influence. A GEO (Generative Engine Optimisation) audit does this through a structured review of how your company and competitors appear across AI-driven search engines (including ChatGPT, Gemini, Perplexity, and Claude), accounting for localised and personalised variations.
It aims to give the most comprehensive possible view – but benchmarking is challenging because of how these algorithms are programmed. Instead of checking backlinks and meta tags, GEO audits examine AI answers, knowledge bases, and brand context. How these search engines respond to queries is based on a much larger, more varied set of data points than traditional search ever used.
While AI search engines now provide some sourcing and can be prompted to give more detail, we cannot fully log how a model arrives at a particular answer. A GEO audit offers the best possible approximation of that cognitive process.
The audit process – How an AI GEO audit works.
An AI GEO audit uses a logical flow to explain what we know about LLMs and business and strategy implementations of that information.
- Baseline analysis: Ask LLMs about your brand, products, and sector. Record mentions, sentiment, and accuracy. Example inquiries include: “What is [Firm X] known for?” vs. “What is the best ESG-focused wealth manager in the UK?”
- Competitive benchmarking: Identify which competitors are being cited in answers your firm should dominate.
- Custom testing: Adjust prompts to see how answers change by phrasing and context. This simulates real-world queries.
- Source mapping: Track which third-party sites, reports, or media are feeding the answers (reviews, articles, awards). Find who is influencing the conversation and where it’s possible to incorporate their work into yours or create partnerships.
- Gap analysis: Where can this be shifted? Prioritise owned sites you can quickly update, including your website and publishing platforms that you control. Other places may be harder to influence, including major publications that only infrequently write about a particular topic.
The full audit should tell a clear, actionable story about your presence online.
What to do with an audit: results at the speed of AI
The results of a good audit will create a blueprint on how to build increased visibility in the short and long term. We recommended creating a definitive plan with benchmarks for action.
Here are several pieces that may be part of an action plan:
- Create authoritative content: Use original research, expert commentary, and clear explanations. The E-E-A-T principles, a holdover from SEO, still very much apply: Content should demonstrate first-hand experience, the source should have a recognised level of knowledge or skill and the website should be seen as a go-to source while seeming safe, reliable, and factually accurate
- Adjust priority media: Press coverage, awards, and high-authority backlinks are increasingly cited by AI search engines. But some trade media have hidden themselves behind serious paywalls or only provide information through newsletters. Re-examine a target list of publications to prioritise those with the most relevant reach.
- Refresh your owned assets: Everything already published should be re-examined for this new age: FAQs, service pages, and product comparisons should be structured semantically for AI parsing. Create a list of “must update” assets to tackle within the next 30 days.
- Target “fan-out” queries: Develop topic clusters that answer tangential but valuable questions where smaller brands can get outsized proportions of search engine volume. The long tail is handy, as typically the most serious and qualified buyers have more nuanced inquiries.
- New connections: Forget turf wars between social, content and comms teams. Treat AI discoverability as a joint responsibility of the broader marcomms team and find ways to ensure collective accountability.
Conclusion
A year from now, AI systems will define your brand before a prospect ever reaches your website. If you’re not actively shaping that definition today, you’re leaving the job to algorithms trained on whatever information they can find, accurate or not.
While overall site traffic may fall in the AI search era, the visitors who do arrive are far more valuable. Ahrefs data shows AI search users convert 23 times better than traditional organic visitors.
The brands that act now will set the standard for how their expertise is understood and surfaced. Everyone else will be trying to rewrite history.
Read this and want your own audit? Get in touch with our cross-disciplinary team to get the process started and stay ahead of your peers.
