EVERYONE's
USING AI

EVERYONE's
USING AI

We’re using it to do something different.

Most agencies are using AI to do the same work faster. We’re using it to do what we do best, better.

Most agencies are using AI to do the same work faster. We’re using it to do what we do best, better.

It’s not about speed. It’s about sharper thinking, smarter decisions and creating work that performs. AI is allowing us to better interrogate how brands show up, uncover where opportunities are being won and lost, and then turn those insights into strategies that drive revenue, build brand relevance and deliver measurable, lasting growth. 

It’s not about speed. It’s about sharper thinking, smarter decisions and creating work that performs. AI is allowing us to better interrogate how brands show up, uncover where opportunities are being won and lost, and then turn those insights into strategies that drive revenue, build brand relevance and deliver measurable, lasting growth. 

Ai Fame Tracker™ from PinPoint Media - Does Ai recommend your brand?


AI Fame Tracker™ is PinPoint Media's proprietary AI Recommendation Intelligence platform. We built it because we noticed something: the brands winning in AI search were not necessarily the best brands. They were the best-structured ones. We wanted to fix that.

More purchase decisions now start inside ChatGPT, Perplexity, and Google AI Overviews than most marketing teams realise. The brands AI recommends are the brands customers choose. Everyone else is invisible before the buying journey has even begun. And they do not know it.

We show you exactly where you stand, which competitors are winning the conversation instead of you, and what to do about it.

see your ai fame score

see your ai fame score

Core AI™️

Core AI™We use AI differently to most agencies. Not to do the same work faster. To make better decisions earlier. Core AI™ sits across our paid social, paid search, creative, and strategy work, handling the research, briefing, reporting, and analysis so our team can spend their time on the things that actually move the needle. Ideas, judgement, and the calls that data alone cannot make.

Core AI™️

Core AI™We use AI differently to most agencies. Not to do the same work faster. To make better decisions earlier. Core AI™ sits across our paid social, paid search, creative, and strategy work, handling the research, briefing, reporting, and analysis so our team can spend their time on the things that actually move the needle. Ideas, judgement, and the calls that data alone cannot make.

How We Use AI

Four principles that govern every brief and every output 

How We Use AI

Four principles that govern every brief and every output 

How We Use AI

Four principles that govern every brief and every output 

01

Clarity before generation

We don’t rush to write prompts. We start with the right questions. We use our FAME model to define exactly what we’re trying to achieve before moving to the next step. Because garbage in means garbage out. 

02

Generate less, decide more

Most AI workflows simply create volume. We use it to improve decisions and test more routes early, so fewer things are produced unnecessarily. No extra work cycles or reworking. Just curation and editing. Better work, with less waste. 

03

360° responsible thinking

We’re B Corp certified, so we measure, avoid, reduce and counter negative environmental impacts. And AI is helping us reduce that further. That means fewer shoots. Fewer unused assets. Fewer dead campaigns. AI shouldn’t just do things faster, it should have a positive impact. 

04

Let AI do the graft

We let AI handle the research, briefing, reporting and analysis. Ideas, taste and judgement stay human. We use AI to free up our time, so we can do the bits that build brand fame.

FAQs

01

Do we weight results against LLM usage?

We deliberately do not weight the headline score by model market share. The score measures consistency across AI systems, not popularity of AI systems. If a brand is only recommended by one model but ignored by the others, that is a weaker signal than being consistently recommended across the ecosystem. We can layer audience or market-share weighting on top later, but first we need a stable and comparable measure of whether AI recommends a brand at all.

01

Do we weight results against LLM usage?

We deliberately do not weight the headline score by model market share. The score measures consistency across AI systems, not popularity of AI systems. If a brand is only recommended by one model but ignored by the others, that is a weaker signal than being consistently recommended across the ecosystem. We can layer audience or market-share weighting on top later, but first we need a stable and comparable measure of whether AI recommends a brand at all.

02

Are competitors who are not listed tracked?

Yes. The platform captures all brands surfaced during testing, including competitors that were not pre-configured. However, only brands that exceed predefined visibility thresholds are surfaced within dashboards, pillar analysis and reports. This prevents reports becoming cluttered with one-off mentions while ensuring emerging competitors are still identified and monitored.

02

Are competitors who are not listed tracked?

Yes. The platform captures all brands surfaced during testing, including competitors that were not pre-configured. However, only brands that exceed predefined visibility thresholds are surfaced within dashboards, pillar analysis and reports. This prevents reports becoming cluttered with one-off mentions while ensuring emerging competitors are still identified and monitored.

03

Doesn’t running thousands of prompts about a brand train the model to think that brand is important?

No. Foundation models are trained on trillions of tokens and updated by the provider on a controlled release cycle. Individual API queries do not alter model weights. Our prompt volume is insignificant relative to the scale of training data. In practical terms, it is a teaspoon in the Atlantic. We have repeatedly tested for contamination effects and found no evidence that measurement activity influences recommendation outcomes.

03

Doesn’t running thousands of prompts about a brand train the model to think that brand is important?

No. Foundation models are trained on trillions of tokens and updated by the provider on a controlled release cycle. Individual API queries do not alter model weights. Our prompt volume is insignificant relative to the scale of training data. In practical terms, it is a teaspoon in the Atlantic. We have repeatedly tested for contamination effects and found no evidence that measurement activity influences recommendation outcomes.

04

How do you stop one prompt from biasing the next?

Every prompt is executed as a fresh, stateless (API) call. There is no conversation history, no shared chat memory and no carry-over context between prompts. Account-level memory features are disabled and we do not reuse user identities or behavioural histories. Each test is therefore conducted in isolation, ensuring results reflect the model’s underlying knowledge rather than previous interactions.

04

How do you stop one prompt from biasing the next?

Every prompt is executed as a fresh, stateless (API) call. There is no conversation history, no shared chat memory and no carry-over context between prompts. Account-level memory features are disabled and we do not reuse user identities or behavioural histories. Each test is therefore conducted in isolation, ensuring results reflect the model’s underlying knowledge rather than previous interactions.

05

What about Perplexity caching your queries or shaping results based on demand signals?

This is one of the key risks we actively monitor. To minimise the possibility of provider-side caching or query pattern effects, we rotate prompt phrasing, vary prompt ordering, randomise execution timing and run every measurement across four independent model families. If one provider were introducing systematic bias, we would expect divergence from the other providers. Our variance and drift monitoring is specifically designed to identify those patterns.

05

What about Perplexity caching your queries or shaping results based on demand signals?

This is one of the key risks we actively monitor. To minimise the possibility of provider-side caching or query pattern effects, we rotate prompt phrasing, vary prompt ordering, randomise execution timing and run every measurement across four independent model families. If one provider were introducing systematic bias, we would expect divergence from the other providers. Our variance and drift monitoring is specifically designed to identify those patterns.

06

How do you know your own testing isn’t influencing the score?

We run contamination testing. Matched accounts are used to execute identical prompt panels, with one account remaining effectively “cold” and another carrying a substantial history of prior measurement activity. Results are compared across hundreds of prompts and multiple model families using confidence interval analysis. To date, observed differences remain well within normal statistical variation. This gives us confidence that the act of measurement is not materially affecting the outcome being measured.

06

How do you know your own testing isn’t influencing the score?

We run contamination testing. Matched accounts are used to execute identical prompt panels, with one account remaining effectively “cold” and another carrying a substantial history of prior measurement activity. Results are compared across hundreds of prompts and multiple model families using confidence interval analysis. To date, observed differences remain well within normal statistical variation. This gives us confidence that the act of measurement is not materially affecting the outcome being measured.

07

What is the biggest source of noise in the scores?

The largest source of variation is the AI providers themselves. OpenAI, Anthropic, Google and Perplexity update their models regularly. Those changes can shift recommendation patterns significantly, often far more than any measurement methodology. That is precisely why we measure across multiple model families rather than relying on a single provider. Every score should be viewed as a directional measure with a confidence range, not a false-precision number. The goal is to identify meaningful trends and competitive movements, not pretend AI systems are perfectly stable.

07

What is the biggest source of noise in the scores?

The largest source of variation is the AI providers themselves. OpenAI, Anthropic, Google and Perplexity update their models regularly. Those changes can shift recommendation patterns significantly, often far more than any measurement methodology. That is precisely why we measure across multiple model families rather than relying on a single provider. Every score should be viewed as a directional measure with a confidence range, not a false-precision number. The goal is to identify meaningful trends and competitive movements, not pretend AI systems are perfectly stable.

08

Why not just measure ChatGPT?

Because no single AI system represents the entire recommendation ecosystem. Different models have different training data, retrieval systems, ranking logic and citation behaviours. A brand may perform strongly in one model and poorly in another. Measuring across multiple providers gives a more robust view of AI recommendation visibility and reduces the risk of making decisions based on the quirks of a single platform. The objective is not to understand ChatGPT. The objective is to understand how AI, collectively recommends brands.

08

Why not just measure ChatGPT?

Because no single AI system represents the entire recommendation ecosystem. Different models have different training data, retrieval systems, ranking logic and citation behaviours. A brand may perform strongly in one model and poorly in another. Measuring across multiple providers gives a more robust view of AI recommendation visibility and reduces the risk of making decisions based on the quirks of a single platform. The objective is not to understand ChatGPT. The objective is to understand how AI, collectively recommends brands.