How to audit your AI visibility: a step-by-step guide
I’m going to walk you through exactly how I’d audit a brand’s AI visibility from scratch as an AI visibility expert. Not the theory — the actual process, the tools I’m using right now (April 2026) and what to do with what you find.
For the purposes of this article, I’m using Libbey, a glassware brand I worked for in the pre-Covid era because I have some institutional knowledge of the brand so I can decipher the data in a deeper way. What we will find is a significant gap between what their marketing says and what AI says about them (their brand identity).
To be transparent, the tools I reference in this guide are what I’m using at the time I’m writing it. This space is moving fast enough that I’ve already swapped out parts of my toolkit twice in the past five months. I’m going to be upfront about all of this with you because I believe that transparency is actually the point — everyone in this space is figuring this out in real time.
What you'll learn in this guide
Step 1: Find out what AI actually thinks about your brand — how to run a prompt audit across ChatGPT, Perplexity and Gemini and what grounding searches reveal that traditional keyword tools don't.
Step 2: Audit your site for AI extractability — how to assess whether your existing content is structured for AI to extract and cite, and what to fix before you create anything new.
Step 3: Map where you exist outside your own site — the full source pool AI pulls from when building your brand identity, and how to find your gaps.
Step 4: Prioritize by platform based on your audience — how each AI platform decides what to cite and how to make practical prioritization decisions for your brand.
Step 5: Build or optimize content for the gaps — where to expand your presence, how to approach Reddit, YouTube and third-party directories as AI visibility channels.
Step 6: Set up monitoring so you're not flying blind — the tools I'm actually using in April 2026 and how to build a monitoring cadence that tells you when things change.
STEP 1: FIND OUT WHAT AI ACTUALLY THINKS ABOUT YOUR BRAND
Do this before you touch ANY content.
Why? AI doesn’t serve up your brand the way Google does. It isn’t just crawling your site, reading the positioning you painstakingly wrote and got approved by leadership. It forms an identity for your brand based on what the ENTIRE internet has decided you are. Product reviews, community forums, Reddit threads, YouTube comments — content you didn’t write, didn’t approve and in most cases have never even read. Scary, right?
That identity gets built whether you participate in the process or not and it matters a heck of a lot more than your homepage copy does when it comes to AI.
I ran an audience intelligence audit on Libbey a while back and what AI had decided about them had almost nothing to do with the brand narrative on their website (one that I actively participated in and wrote!). Their AI identity was built around “budget-friendly” and “dishwasher safe.” Not the timeless design and entertaining vibe the brand pushes in copy and creative assets — which is really unfortunate because I was a part of multiple photoshoots that required significant investment, where that’s literally all we portrayed — and those photos are still used across the website and product listings. None of the language the marketing team (or I) wrote was included in the identity. But what was included is the stuff their customers have been saying in reviews and forums for years — and the brand is 183 years old so...that’s a lot of material to go off of.
This problem is not only applicable to Libbey, it’s impacting every brand.
How to run a basic prompt audit
Open ChatGPT, Perplexity and Gemini. To run your prompt audit, enter the same set of queries across all three and document what comes back. You’re looking for:
How does each platform describe your brand?
What category do they put you in?
What words do they use that you would never use yourself?
Are there factual errors in how they describe you?
Are you mentioned at all for the queries your customers are actually asking?
That last question is the one that can either make or break your day. A brand can have invested years in a category and be completely invisible to AI for the searches related to it. Talk about a 2026 marketer’s nightmare.
What are grounding searches and why do they matter
When you run a prompt audit manually, as I laid out above, you’re seeing the surface of AI search. What you can’t see without a dedicated tool are grounding searches. These are not search terms that any human would type in, so they haven’t been on the radar of your content strategy, but now they’re incredibly important because they directly determine whether AI cites you.
When I ran a Spyglasses audit, a dedicated AI visibility tool, for a client alongside their Semrush data, Semrush showed them leading their category at 85% share of voice. Spyglasses showed 0% share of voice on the single most important query for their business. How are the two so different?
Well, it’s not that Semrush is wrong, it’s just measuring something different than whether AI understands your brand for the specific problem you exist to solve. Spyglasses surfaced 31 grounding searches my client was absent from entirely. Semrush showed none of them.
Those searches are now your starting point — not your keyword rankings or traffic.
Read more: What your AI visibility score can’t tell you
STEP 2: AUDIT YOUR SITE FOR AI EXTRACTABILITY (NOT JUST SEO)
After you learn your brand is not represented in AI accurately, the most important thing to do is to NOT create NEW content. The first thing you have to do is more boring — you need to audit the current content. Because new content on top of a broken foundation is a little bit like polishing a turd, or putting lipstick on a pig (no offense — your content is probably great in the traditional sense, but like I said, we have to rewrite the script).
What AI extractability actually means
When AI reads your site, it looks for self-contained answer blocks it can pull and cite. Unfortunately, the human-centered narrative that takes three paragraphs to make a single point is hard for AI to use, as depressing as that sounds for copywriters. A 50-100 word passage that completely answers one specific question is exactly what it’s looking for.
I didn’t say you or your team would like every step of this process, it’s probably going to piss some people off, actually.
You have to think about the fact that, when someone asks ChatGPT a question, it isn’t going to summarize your entire 2,000-word blog post. It’s going to find the specific paragraph that answers the specific question and pull that. If your content isn’t written in extractable units, AI has nothing clean to grab.
Here’s what AI-extractable content actually requires:
Q&A structure and definition blocks
Every key concept on your site should have a plain-language definition written as a self-contained 50-100 word passage. Not in paragraph three, not speckled across sections — you need one question and one answer.
For example, scroll up in this article and notice how I defined grounding searches as “the queries AI platforms run in the background to retrieve information before generating a response.” That’s a definition block — one concept, plain-language answer, fully self-contained. The header is “What are grounding searches and why do they matter” directly above it. That, my friends, is a heading plus a definition written in exactly the structure AI looks for.
FAQs
You’re going to start adding more FAQs than feels comfortable. Every question your customer has ever asked, every objection they’ve ever raised, every comparison they’ve ever made. If it’s a question that exists somewhere on the internet about your category, it belongs in your FAQ.
Clear H2 and H3 headings that describe what each section actually answers
Another part that is going to make copywriters scream internally (or out loud, I don’t judge). The headers need to be descriptive, not clever or branded. “How AI answer engines decide what to cite” is a better heading than “The intelligence behind our approach” if you want AI to find and use that section.
Schema markup
All the schema — FAQPage, Article, Organization and HowTo schema, depending on your content type. This is the structured data layer that helps AI systems understand and trust your content. How you implement it depends on your CMS — this is one you want to tag-team with your developer or SEO specialist and make sure you’re maximizing it for AI optimizations, which might mean more schemas than you’d typically add per page.
Entity consistency
Your brand name, description and services need to read the same way across every property. That means your website, LinkedIn company page, Google Business Profile and directory listings need to all match as closely as possible. At least the main points you mention, since we know each platform has different character counts and best practices. This is important because AI cross-references all of them and inconsistency won’t help.
Where Semrush fits and where it stops
If you already have the Semrush toolkit, with the info we’ve reviewed in this article, you will hopefully be able to look at the information it’s giving you in more granularity now. Take note of the side-by-side view of your AI visibility and traditional SEO rankings.
However, be mindful that it measures AI share of voice based on traditional search so it won’t show the grounding searches that AI uses to understand your brand for the specific problem(s) you exist to solve.
Spyglasses
A dedicated AI visibility audit tool, Spyglasses provides a point-in-time audit with the complete view of how AI understands your brand, what grounding searches you’re absent from and how you compare to competitors. This info is gold because it becomes your list of what needs fixing.
This is the tool that surfaced 31 grounding searches my client was absent from entirely — searches that Semrush doesn’t show.
It’s also what showed 0% share of voice on their most important query while Semrush was showing 85%. I’m no mathematician, but I know that’s not anything close to the same equation.
If you want my top recommendation for you — but I know only less than 10% will do this part today? Run a Spyglasses audit before you do anything else in this guide. It’s worth the additional fee to your stack because the findings directly inform every step that follows — what to fix on your site, where you need to build presence, which platforms matter most for your specific brand. The audit is the literal map.
How to think about monitoring going forward
The Spyglasses visibility report you run today becomes your AI visibility baseline. Then set up a cadence of running a new audit — when you’ve made significant changes like new content is published, the site is restructured or you’ve expanded your presence to new platforms.
Continue to use Semrush for your traditional SEO, and if you have the AI toolkit, continue to monitor the report and any new capabilities.
Nothing about this is a set-it-and-forget-it process. AI platforms are constantly evolving — updating their training data, changing their source weighting and shifting their citation behavior. The marketers that build a monitoring foundation now, and evolve it with their knowledge and strategy, will become the new AI visibility experts because they’ll know when something changes and why. The marketers that don’t become students to this new area of search will be stripped of their “search expert” title in a few years (or sooner).
STEP 3: MAP WHERE YOU EXIST OUTSIDE YOUR OWN SITE
What is an AI brand identity?
This is the step that seems weird when you compare it to your current content strategy. Your main focus has been driving traffic to your website, optimizing your pages, building your blog — and now I’m going to tell you to do something you’ve probably never thought you’d do.
AI doesn’t care about your conversion funnel.
When AI generates a response about your brand or your category, it isn’t going to your website to get the answer. It’s pulling from everywhere it can find you, whether you’re speaking or it’s your customer.
Reddit and community forums
This is the one that surprises people most when they first start researching AI visibility. Reddit accounts for 46.7% of Perplexity citations. If your customers are talking about your category on Reddit, Houzz, industry-specific forums or professional communities, that is the content actively shaping how AI describes you. And you didn’t write one word of it.
Product reviews
Depending on your category, Google reviews, G2, Capterra, Trustpilot and Amazon reviews greatly impact your AI identity. Remember the Libbey example from earlier? Their AI identity was built almost entirely from what customers wrote in reviews. It wasn’t taken from the brand copy or the photoshoots.
YouTube
We are all underutilizing YouTube right now. An analysis of AI-cited YouTube content found that 40% of cited videos have fewer than 1,000 views. Video views and engagement does not predict citation — it’s a marketer’s dream!
A well-structured, long-form video that clearly answers a question your customer is asking will get cited by AI regardless of the views. Gemini in particular favors YouTube content because it’s a Google property.
Wikipedia
If your brand or category has a Wikipedia page, AI treats it as a high-trust source. Crazy, right?
Press mentions and news coverage
Third-party sources offer credibility signals that AI weighs heavily, especially ChatGPT which draws from training data where journalism carries authority. So dust off that PR hat, or bring in the Communications team to strategize opportunities to increase your AI visibility.
Podcasts
Optimizing podcast transcripts for AI to read and cite. If you’re already recording them, this is the easiest addition to your strategy. If you’re not, might be time to buy that mic.
Substack and newsletter content
Subscription content is an emerging cited source, particularly on Perplexity as independent expert voices continue to grow and gain authority. LinkedIn newsletters contribute to your visibility but independent publishing platforms like Substack carry more citation weight because AI treats them as standalone publications. This is not a content type I have experience with, but have been thinking of digging into it for my own brand.
Industry directories and listings
If applicable, directories like Clutch, UpCity, ThomasNet for manufacturing and Capterra for software are heavily weighted third-party signals for the grounding queries that directly feed AI citations.
Where does your brand exist outside your own website?
Go through each one and check. Search your brand name on Reddit. Look at what your reviews actually say (not just the star rating). Think about your YouTube presence and whether any of it is actually intentional and not just uploaded as a campaign afterthought. Find out if you’re listed on the directories that matter for your category.
How to utilize your external presence audit for your strategy
Going back to Libbey — a 183-year-old brand with reviews across platforms that date back to basically the start of the internet, historical mentions across community forums, and a widespread product listing presence (and the accompanying reviews) across platforms. Their AI identity isn’t built from their website when the internet has years of customer voice to draw from. The “budget-friendly” and “dishwasher safe” identity didn’t come from nowhere — it’s from the list above, which was built up over decades, without the marketing department’s involvement.
That is where AI brand identities come from. While your brand may not be as extreme (because you’re probably not 183 years old), the truth is, you probably still have some significant gaps.
STEP 4: PRIORITIZE BY PLATFORM BASED ON YOUR AUDIENCE
What is AI share of voice?
“Optimizing for AI” isn’t one thing. Because nothing is ever easy in love, each platform has a different architecture, pulling from different source pools with different weight signals. What gets you cited on Perplexity might do nothing for your Gemini visibility.
An analysis of 6.8 million citations found that only 11% of domains receive citations from both ChatGPT and Perplexity. Add Google AI Overviews into the mix and the overlap shrinks even further.
Here’s what each one actually prioritizes:
Perplexity
Reddit is the dominant citation source at 46.7% of citations. Perplexity searches the live web for every query, which means freshness matters — new content can appear in Perplexity citations within days of publishing. It also cites roughly 13 sources per answer on average, which makes it the most generous platform for brands that aren’t household names.
If you publish consistently and your content is well-structured, Perplexity is likely where you’ll see traction first.
ChatGPT
Probably the most mainstream platform at this time, ChatGPT draws heavily from Wikipedia, Reddit and established publications. For real-time queries it uses Bing-powered search.
Training data is essentially the internet snapshot that ChatGPT learned from before it was released — think of it as everything that existed and was publicly crawlable up to a certain point in time, weighted toward sources AI systems treat as authoritative, including Wikipedia, Reddit, publications and review platforms. ChatGPT draws from that snapshot when it talks about your brand, which is why what people have written about you historically matters as much as what you’re publishing today.
Third-party validation matters on ChatGPT more than on any other platform, so if your brand isn’t in any conversations outside your own website, the training data has limited material to draw from. Think reviews, press mentions and directory listings — big money content in AI.
Gemini
Favors YouTube, because duh, they look out for their own family. Google owns YouTube, so if your brand has video content and optimized descriptions that answer questions your customers ask, Gemini is the platform most likely to surface it. Gemini also draws from its other owned platforms, including Google-indexed content and Google Business Profile data.
Google AI Overviews and AI Mode
This is the most SEO-impacted AI presence because it requires your content to rank in the top 20 organic Google results before being considered for AI citation. If you’re not ranking organically, Google AI won’t cite you. This is the one place where your existing SEO investment has the most direct transfer to AI visibility.
How to prioritize AI platforms for your brand
You should be thinking about them all. In my opinion, it’s too early in the AI search game to be confident your audience isn’t using all of them. I know that’s not the answer anyone wants because it means more steps in the content optimization process, more platforms, more analytics to follow.
That said, you can make some practical decisions based on your current content and your resources. A brand with an existing YouTube presence has a clear Gemini opportunity sitting there, so put the bow on that process by setting up your SOP for description optimization. A brand investing heavily in SEO already has infrastructure that transfers directly to Google AI Overviews. You might as well get the platforms you’re 80% of the way optimized for to 100%.
STEP 5: BUILD OR OPTIMIZE CONTENT FOR THE GAPS
Why you should audit existing content before creating anything new
Please, listen to me when I say, do not publish new content before you update your current. Doing so is like publishing an article on a sinking foundation. Go back to your site audit from Step 2 and work through the content, adding definition blocks and FAQs, fix the headings, add additional schema, and ensure you have entity consistency across every property.
This is the non-shiny work, and depending on how many pages your website is, it may require a lot of time and resources. But every day you leave an unstructured page live is another day AI has nothing clean to extract from it.
Then expand your presence outside your site
Once your foundation is solid, you build outward. Based on your external presence audit from Step 3, identify the two or three sources where you’re most absent and where your audience is most active.
Don’t try to be everywhere at once.
You can’t go into the office on a Monday and all of a sudden expand your content creation across new platforms. But, there are a few worth prioritizing:
YouTube
Starting to create long-form video content that directly answers customer questions is probably one of the biggest untapped AI visibility opportunities; however, I know that’s a heavy lift. And since structure matters more than production value, meaning a video with a clear title, a keyword-rich description and content that methodically answers one question, an easy way to enter this platform is to utilize the short-form videos created for social and publish them as a YouTube video or YouTube Short (less than 60 seconds). Optimized content will outperform a polished brand film for AI citation purposes every single time.
I know it seems crazy! Unless you’re a legit Reddit user, you probably only end up on the platform when you’re doing random research. Ding ding ding!!!
If your customers are having conversations about your category on Reddit and you’re not mentioned in the threads or monitoring the conversation, you have a huge hole in your boat. The difficult part about this is, it doesn’t mean your typical brand posts on a social platform — it means showing up in the communities where your audience already is and contributing genuinely. Reddit users can sniff out marketing a mile away, so this has to be done delicately. Only a small category of brands can show up branded. For example, doing an audit for an automotive parts brand, their audience was asking Reddit for the part number of that brand they needed to fix something. THIS is the type of conversation a brand can step into and answer. But the typical brand needs more of an influencer-esque strategy — think advocates answering questions your audience has, doing so transparently but genuinely. This includes everything from CPG brands to healthcare and MedTech, which isn’t just simply launching another campaign. But the brands that figure how to create an authentic Reddit presence in 2026 are going to have a citation advantage that is very hard to close later.
Third-party directories and review platforms
The unsexy platforms really are king in AI, like Clutch, UpCity, G2, Capterra and ThomasNet, depending on your category. Having another third-party platform full of detailed reviews that use specific service language makes a huge difference.
Press and earned media
Clutch your pearls because one industry publication mention can carry more ChatGPT citation weight than a hundred blog posts on your own domain. If PR hasn’t been part of your strategy, AI visibility is a new reason to call in the Comms team.
Why Reddit is now a legitimate content strategy channel
None of us thought we’d be building Reddit strategies two years ago, but those of us who start truly integrating it into the strategy are going to be thankful in a year.
Choosing the right mix of channels for your brand to start investing in depends entirely on your category, audience and where they actually spend time. There isn’t a playbook for this yet, and if anyone tries to sell you one, it’ll be outdated within a month because the landscape is changing so fast.
Use the audience intelligence work from Step 1 to inform every decision you make in this step. What your customers are actually saying, where they’re saying it and what questions they’re asking — that’s your AI visibility content strategy.
STEP 6: SET UP MONITORING SO YOU’RE NOT FLYING BLIND
Most brands are completely blind to what AI is saying about them, because the people who are expected to provide that answer are just people like you and I, over here trying to navigate this marketing landscape.
This space is new, so that means the tools are too. My own toolkit has changed twice in the past five months. Here’s where I would recommend you start today, April 2026.
Semrush AI Toolkit
If you already have the Semrush toolkit, with the info we’ve reviewed in this article, you will hopefully be able to look at the information it’s giving you in more granularity now. Take note of the side-by-side view of your AI visibility and traditional SEO rankings.
However, be mindful that it measures AI share of voice based on traditional search so it won’t show the grounding searches that AI uses to understand your brand for the specific problem(s) you exist to solve.
Spyglasses
A dedicated AI visibility audit tool, Spyglasses provides a point-in-time audit with the complete view of how AI understands your brand, what grounding searches you’re absent from and how you compare to competitors. This info is gold because it becomes your list of what needs fixing.
This is the tool that surfaced 31 grounding searches my client was absent from entirely — searches that Semrush doesn’t show.
It’s also what showed 0% share of voice on their most important query while Semrush was showing 85%. I’m no mathematician, but I know that’s not anything close to the same equation.
If you want my top recommendation — but I know only less than 10% will do this part today? Run a Spyglasses audit before you do anything else in this guide. It’s worth the additional fee to your stack because the findings directly inform every step that follows — what to fix on your site, where you need to build presence, which platforms matter most for your specific brand. The audit is the literal map.
How to build an AI visibility monitoring cadence
The Spyglasses visibility report you run today becomes your AI visibility baseline. Then set up a cadence of running a new audit — when you’ve made significant changes like new content is published, the site is restructured or you’ve expanded your presence to new platforms.
Continue to use Semrush for your traditional SEO, and if you have the AI toolkit, continue to monitor the report and any new capabilities.
Nothing about this is a set-it-and-forget-it process. AI platforms are constantly evolving — updating their training data, changing their source weighting and shifting their citation behavior. The marketers that build a monitoring foundation now, and evolve it with their knowledge and strategy, will become the new AI visibility experts because they’ll know when something changes and why. The marketers that don’t become students to this new area of search will be stripped of their “search expert” title in a few years (or sooner).
THE PROCESS THAT TIES IT ALL TOGETHER
Congrats if you’ve made it this far in the process, because I know it’s a bit overwhelming. We’re not talking about a plugin you install or one more line on the checklist. It’s a NEW workstream that runs alongside everything your marketing team is already doing.
When you start implementing these tactics, you need to follow the proper logic.
What you just read is the foundation of what I call the AI Intelligence Loop — a four-step framework that connects audience intelligence, strategic translation, content activation and AI visibility monitoring in a continuous cycle. Each step feeds the next and the last feeds back into the first.
You start by listening.
What is AI saying about your brand?
What are your customers actually saying in the places you didn’t write?
That’s your audience intelligence — the input that helps you make every other decision.
You translate what you find into the strategy — not what your next blog topic is, rather the picture of what needs to change, what needs to be built and where the gaps are between your brand identity and your AI identity.
You activate — fix the foundation, expand your presence and create content that AI can actually extract and use. You show up in the places AI looks, not just the places your marketing department has always looked.
And then you monitor — more often than your monthly check-in for the analytics report. AI platforms are constantly evolving, citation behavior is shifting and the competitive landscape in AI search can shift within a week. The monitoring loop feeds back into the intelligence step and the process starts again.
The brands that win AI visibility in the next few years won’t be the ones that did one audit and published ten new blog posts. They’ll be the ones that got in the trenches and built an ongoing function — a strategy that evolves WITH the landscape because someone on the team is paying attention.
That’s the work of AI visibility. It is exhausting, unless you love a good ol’ rabbit hole to go down like I do.
My biggest piece of advice is to figure it out now, while most of your peers are still treating AI visibility as a future problem. That’s an advantage for your brand, and for you in the job market.
Now go on grasshoppers, go optimize!
Frequently asked questions about AI visibility audits
-
An AI visibility audit is the process of testing how AI platforms like ChatGPT, Perplexity and Gemini describe your brand, what sources they cite and whether your brand appears in the answers your customers are actually seeing. It goes beyond traditional SEO metrics to reveal how AI understands your brand identity and where the gaps are between what you say about yourself and what AI says about you.
-
Run a basic prompt audit. Open ChatGPT, Perplexity and Gemini, search your brand name and the core problem your product or service solves, and document what comes back. Look for incorrect descriptions, missing services, outdated positioning or instances where your brand isn't mentioned at all. For a deeper picture, a dedicated tool like Spyglasses will surface the grounding searches AI is running about your category that you can't see manually.
-
Grounding searches are the queries AI platforms run in the background before generating a response. They are not search terms any human would type — they are the internal queries AI uses to retrieve and verify information. They directly determine whether AI cites your brand, and they don't appear in any traditional keyword research tool.
-
Traditional SEO optimizes your content to rank in search results so humans can find and click through to your site. AI visibility optimizes your content and your broader digital presence so AI platforms can extract, understand and cite your brand when generating answers. SEO is about ranking in a list, while AI visibility is about being referred and/or cited in one answer. Both matter and they work best together, but they require different strategies and different measurement tools.
-
Run a full audit at the start of your AI visibility strategy to establish a baseline. After that, run one any time you make significant changes — a site restructure, a major content push, an expansion to new platforms or a repositioning of your brand. AI platforms update their training data and citation behavior regularly, so monitoring on a quarterly cadence at minimum is recommended.
-
Start with the platform your audience uses most. If you don't know that yet, start with Perplexity — it's the most generous platform for brands that aren't household names, it cites roughly 13 sources per answer and new content can appear in citations within days of publishing. From there, build toward ChatGPT (third-party validation and reviews) and Gemini (YouTube and Google-indexed content). Google AI Overviews requires top-20 organic ranking first, so traditional SEO work feeds that one directly.
-
No, the first step is auditing and restructuring what you already have, not deleting it. Add definition blocks, expand your FAQs, fix your headings to be descriptive rather than clever and add schema markup. Only after your existing foundation is solid should you create new content to fill the gaps your audit surfaces.
-
At minimum, start with manual prompt audits across ChatGPT, Perplexity and Gemini — this requires no tool and gives you a surface-level picture immediately. If you have Semrush and there’s less internal friction to add the AI Toolkit over adding a new tool entirely, do that for a baseline share of voice score alongside your traditional SEO data. For the full picture including grounding searches and competitive gap analysis, Spyglasses is the dedicated AI visibility audit tool I use with clients.
