AI search has changed how audiences discover brands. Whether you’re being recommended in an LLM answer or highlighted in Google’s AI Overviews, concise AI summaries are fast becoming a key touchpoint for directing new audiences towards your brand.
But how do you actually monitor these responses, let alone get a complete picture of your brand’s AI narrative? You can’t influence what you don’t measure, so here are our top tips for uncovering exactly what AI is saying about your brand.
Start by researching the topics your buyers care about
If you assume everyone aware of your brand already has an idea of what you do, the best place to start your research is with your key services or product ranges, treating each one as an individual topic. Search these terms on Google and LLMs and review for any appearance of your brand.
If you sell to several audiences, test each one separately. A business buyer asks different questions to a consumer, and gets different answers. In most cases a different audience means a different topic altogether.
Cover three types of prompt
- Category questions, where the buyer hasn’t named any brand
- Comparisons, where the buyer weighs you against competitors
- Brand questions, where the buyer asks about you by name
For a digital marketing agency, a starting set might look like this:
- What are the key components of an effective digital marketing strategy?
- How can I develop a comprehensive digital marketing plan for my business?
- How does an integrated approach improve online marketing results?
- What tools are most useful for analysing the digital marketing landscape?
- How does [your brand] compare with [competitor] for B2B paid media?
- What do clients say about working with [your brand]?
Check the platforms your buyers use
ChatGPT is the obvious place to start. Claude and Gemini, including Google’s AI Overviews and AI Mode, are now just as important. Perplexity is more niche, but it matters in some industries.
Checking by hand works well for top-level questions about your brand or key services. To go further, a GEO tool such as Semrush can run far more prompts across more platforms.
Even with a tool, read a sample of the responses yourself. The overall picture is only useful if you understand what sits behind it.
AI answers vary between platforms, between models and between runs. The models themselves are changing fast. One check is a snapshot, not a verdict.
To make manual checks comparable over time:
- Use the same prompts every time
- Check Google’s AI Overviews in an incognito window
- Use temporary or private chat modes in other AI tools, so your history doesn’t shape the answer
- Log every result in one shared spreadsheet
Record the results (and keep an eye out for patterns)
A simple scorecard keeps every check comparable. Use one row per prompt, per platform, per check.
| Date | Platform | Prompt | Brand mention | Top brands names | Sentiment | Full response |
| Month and year | e.g. ChatGPT | The exact prompt used | Yes or no | The top three to five brands in the answer | Positive, neutral, negative or inaccurate | Paste the whole answer |
Sort sentiment into a few fixed categories so you can compare month to month. Keep the full response rather than a summary. You’ll need it when you trace where an answer came from.
One negative answer isn’t necessarily a problem. What matters is whether the same claim shows up across several platforms over several months.
Scale matters too. One or two critical reviews are a big deal for a young business and minor for an established one.
When a claim keeps recurring, go to the sources. Most AI tools show where an answer came from, so visit those pages and judge whether they’re credible. A recognised source generating a lot of noise about your business is worth acting on.
Schools show how this works in practice. Few businesses are as personal, and some parents will always raise concerns online. Unless those posts point to a wider issue, chasing them rarely helps.
Some forums will remove commentary that’s false, which is worth knowing. Otherwise, the better use of time is building visibility and positive sentiment elsewhere.
Know when to move beyond manual checks
Manual checks are the place to start. Your marketing lead is best placed to own them, either as part of day-to-day work or as a monthly routine.
Treat them as a first step. GEO tools are relatively expensive, so first find out whether there’s much about your brand out there to track. Once there is, a tool or a GEO specialist will track it far more reliably than manual checks.
Look at AI referral traffic in GA4 alongside your checks. It shows how many people reach your site from AI tools. Much of the research now happens inside the AI answer itself, so referral traffic will only ever show part of the picture.
Understand and control your brand’s AI narrative with Loom
Our AI sentiment strategy article covers why that matters and how to shape it.
If your checks show negative sentiment or very little visibility, that’s the point to bring in a GEO specialist. They know what normal looks like and can help you shift the narrative. Our AI sentiment strategy article sets out the steps.
GEO doesn’t replace SEO. A technically sound, well-structured website with clear, useful content is still the foundation for showing up in AI answers. Our guide to SEO and GEO for complex businesses covers how the two fit together.
