Why AI Search Visibility Matters (More Than Ever)
Today, you know how important it is to rank in AI search. But with all these AI visibility platforms and new self-proclaimed AI SEO consultants competing for a piece of that sweet, sweet marketing budget, it can be hard to know what’s real and what’s fraudulent. Well, in this blog, I’m going to give you the power to cut through all the noise and actually perform an AI search audit for yourself for free, without the not-so-subtle cost of these AI assistants.
It’s a manual exercise and a bit tedious. But I believe that before spending money on an automated AI visibility tracker, you should understand how large language models think, work, and what makes them valuable lead generation channels.
And the organic channel that’s trending these days is almost always AI search. But before we dive into the audit process, let’s clarify what we’re actually trying to achieve here.
What Are We Actually Trying to Achieve?
In AI search, there are two ways to win.
First, you can be cited as a source. This happens when your URL appears in a footnote or reference in an AI search conversation.
This is a good sign because it means the AI found your content valuable enough to reference. But even better is having your brand name mentioned in the actual answer. And if you can get both a direct mention and a citation, that’s perfection. But here’s the challenge: AI visibility audits are much more difficult than traditional SEO audits.
With traditional search, we have a wealth of data, reliable keyword research tools, and predictable ranking factors in Google Search Console. AI search, on the other hand, is much more volatile, changing in real time. For one thing, AI assistants are generative, meaning no two answers will likely be the same because the large language models inside are generating answers on the fly.
Second, factors like personalisation and user search history will influence how AI assistants display specific answers. As I mentioned earlier, because OpenAI, Perplexity, and Anthropic are private companies, we don’t have access to query databases to see what people are searching for. So, AI search is a bit of a black box.
So, we can’t rely on analytics. Instead, we reverse-engineer visibility by prompting AI ourselves.
Step 1: Pick Your AI Platforms
Thankfully, we can extract insights by prompting it and observing how it gathers information, as well as studying its thought processes. So, for this audit, we’ll focus on at least one of the four major AI platforms. ChatGPT, Google AI Overview, Cloud, or Perplexity. Now, you’re free to start with any one of these, but if you handle all four, you’re actually covering 90% of all conversational AI searches.
Step 2: Craft the Right Queries
Next, because we don’t have access to the actual queries people are searching for, we need to put ourselves in the customer’s shoes and understand what they might be searching for. While traditional SEO focuses more on the top of the funnel with high-volume, low-competition, informative posts to drive traffic, for this exercise, we’re focusing entirely on the middle and bottom of the funnel.
An example of a top-of-the-funnel query might be, “Why do I need an email marketing platform?” If you’ve written content on this topic, you can be cited as a source. However, our goal is to achieve both a citation and a mention by name, which is typically the case with queries at the middle or bottom of the funnel. For example, “What are the best email marketing platforms for small businesses?” This searcher’s intent is obvious.
They know they have a problem that only an email marketing platform can solve, and now they’re in the evaluation phase. I recommend brainstorming two types of queries.
The first is discovery queries without a brand name. This is to test organic appearance. The second is brand-specific queries to test sentiment and accuracy.
For discovery queries, these might include things like,
- “What are the best email marketing platforms for small businesses?”,
- “Which project management tools integrate with Slack and Google Workspace?”,
- “Which CRM software works best for real estate agents for under $50 per month?”,
And for brand-specific queries, you might ask,
- “How does your brand compare to Mailchimp for email marketing?”,
- “What are the advantages and disadvantages of your brand compared to competitors?”,
- and “Is your brand worth the price compared to other options?”
In my experience, this is very beneficial.
Talk to your customer support and sales teams for ideas. These are the people on the front lines, actually talking to customers, understanding their pain points, and how your product or service can solve them. And if you need inspiration, I’ve found AI tools to be very helpful here. I’ve added a prompt in the description of this video that you can paste into your favorite AI assistant, which will study your business and give you some ideas for prompts for your company.
It’s up to you how many prompts you want to test, but I recommend at least five in each category. Once you’re happy with your prompts, it’s time to systematically collect AI responses. Now, there are certainly developer types who would recommend using these companies’ APIs to build custom applications that systematically query each of these keywords.
Step 3: Run the Audit (Manually)
Now, if you feel comfortable creating or coding such a script, go for it. But if you’re non-technical and don’t want to mess with code, you can get the same results in a browser. For each query, I recommend opening each platform in a new tab, making sure you’ve enabled ChatGPT Perplexity and Temporary Cloud Mode, and opening Google in an incognito browser.
There are still significant differences between logged-in and logged-out Chat GPT Chat, free or paid plans, regardless of which model you’re using. All of this adds significant variation to AI responses. So, while it’s not perfect, we want to do our best to isolate our variables and standardize these experiments.
Step 4: Analyze Your Results (With AI’s Help)
Next, create a spreadsheet with columns representing each AI assistant and rows corresponding to each query. Anyway, once you’re ready, simply query each assistant and copy and paste the output into the corresponding cells in your spreadsheet.
Before moving on to the next question, I recommend taking a few seconds to click the Thinking button on ChatGPT. Depending on which AI assistant you’re using, it may have a different name, and see how the AI assistant derived its answers. Similarly, take a moment to hover your mouse over each source mentioned in the answer.
You’ll definitely start to see patterns in the frequently referenced sources. Once you’ve completed your entire spreadsheet, this is where the fun begins. Not only can you manually analyze these responses, which is certainly a useful and helpful exercise, but you can also feed this CSV back into a larger language model to gain powerful insights at scale.
Here are the metrics I recommend looking for and the prompts to get them.
1. Mention Volume
Number one, mention volume. Now, mention volume tracks how often your brand appears in AI responses. This metric helps you understand your overall AI visibility. The prompt to measure mention volume, again, simply takes it from the description. The AI will analyze the CSV of responses and count how many times each brand name appears across all responses. It will also show patterns where certain types of queries generate more mentions for specific brands. Now, of course, all your branded queries will mention your brand because they are branded queries. So this metric is most helpful if you isolate your unbranded queries and have the AI assistant analyze only those.
2. Sentiment Analysis
Number two, sentiment analysis. Sentiment analysis looks at whether AI mentions of your brand are positive, negative, or neutral. This helps you understand not only visibility but also brand perception. To do this, you’ll need to enter both branded and unbranded queries and responses. The sentiment analysis prompt will analyze the underlying sentiment of each brand mention and categorize it as positive, negative, or neutral, along with the context behind it, whether it’s pricing, features, support, etc.
3. Frequent Source Analysis
Number three is regular source analysis. Now, this experiment can reveal certain websites or publications that the AI assistant frequently quotes. These reveal high-authority sources you should target for coverage or compete with. The prompt to uncover frequently cited sources will identify all sources or URLs mentioned in the AI response and rank them based on their source domain, number of citations, type of query cited, and the competitive brands mentioned in each source.
It will also highlight opportunities where your brand can be featured on frequently cited sources or simply outrank them by creating competitive content.
4. Competitive Share of Voice
Number four, we want to determine your competitive share of voice. This is your brand’s relative visibility compared to its main competitors across all AI platforms. This provides context for your performance and identifies which competitors you need to target in AI search.
This prompt will calculate the competitive share of voice by identifying all brands mentioned, calculating each brand’s total mentions as a percentage of all brand mentions, and displaying the share by query type and AI platform. I know it’s a lot, but if you copy and paste the prompt into your CSV, I promise it’ll all make sense.
It will also highlight your biggest competitive threats and opportunities to gain share. Now, when it comes to slicing and dicing this data, there’s no limit. These are just some helpful starting metrics, but you can also have AI analyze your citation vs. mention ratio, query intent performance, and, frankly, anything else you can think of.
I recommend running these analysis prompts through at least two different large language models, such as Cloud and ChatGPT. And yes, be sure to QA and verify the data before showing it to any stakeholders. Trust, but verify. That’s the name of the game when it comes to using AI for data analysis. And once you have these raw insights, you can use AI to format them into a step-by-step action plan.
Step 5: Turn Insights Into Action
Simply paste the action plan prompt into the same chat as your insights or the previous prompts, and it will create a prioritized 90-day action plan with specific tactics to improve your AI search performance. You’ll definitely want to edit this before showing it to anyone else, but in my experience, it’s a really good starting point.
Measure, Iterate, Dominate
If you find the large language model is confusing responses or the data is too much to handle, start by answering one assistant at a time. Now, once you’ve actually taken action on your plan, I recommend waiting about a month and running this entire process again.
Use the same prompts, the same help, and track the same metrics to see if your actions are actually making an impact. So, that wasn’t so bad, right? This manual audit process may seem tedious, but it’s the foundation for understanding how AI search actually works. Once you master this systematic approach, you’ll know how to evaluate any AI tracking tool that comes to market.
And more importantly, you’ll understand exactly what steps to take to improve your visibility. Whether you’re a solopreneur trying to get mentioned alongside big names in the industry or a well-known company looking to maintain its market position in the age of AI SEO, this audit framework gives you the data you need to make informed decisions.
