The Workflow I Use to Track Performance in ChatGPT & Google AI

Tracking your visibility in AI engines isn’t a nice-to-have anymore - it’s a must. Hi, I’m Anastasia Kotsiubynska, and I lead the SEO team at SE Ranking. I’ve spent the last year testing how AI search engines like ChatGPT, Google’s AI Overviews, and AI Mode actually surface content — and how we can track visibility there even without perfect data. Okay, let’s get into some numbers. AI platforms now drive about 0.15% of all global traffic. Sounds tiny? Maybe - but that’s a 7x jump since 2024. So yeah, the overall share is still tiny - but the growth and engagement? Massive. Your traffic sources are probably shifting - whether you know it or not. But here comes the real problem. As an industry, we’re flying half-blind. Google Search Console doesn’t show proper data for AI Overviews or AI Mode. We have zero visibility into the actual queries or prompts from AI engines. And the big question still hangs in the

air — how can we actually fit AI traffic into your performance metrics and measure what it’s worth in the bigger business picture? In this video, I’ll show you a few practical ways I’ve been using myself to monitor visibility in ChatGPT, Google’s AI Overviews and AI Mode, and other AI engines: what to look for, which tools can help so far, and how to use AI itself to speed up some of your SEO work - including pulling your AI visibility data directly into ChatGPT or Claude through MCP. And to do that right, we first need to understand the landscape - where all this AI traffic is coming from. While there are a lot of AI engines, they’re not made equal. Google’s own AI search features, like AI Overviews that is already present in around 30-40% of queries in the US and is very common around the world, AI Mode, first released in May for the U.S. and probably

becoming a default Google’s option one day. At first glance, AI Mode and AI Overviews look like part of the same feature set. But when we analyzed 10K keywords, the URL overlap was shockingly small - only around 10%. In fact, more than half the time, they barely shared any URLs at all, which means we still need to track and analyze our presence there separately. But the issue is that you don’t have first-party tools and data sources to understand your performance in AIO and AIM. Google doesn’t make it easy —Search Console shows all organic clicks from Google, but there’s no way to separate this traffic. Stick around — I’ll show you how to track your AI visibility right now using the available data, SE Ranking’s tools and a few smart workarounds. Next, there are standalone AI platforms. And here, ChatGPT is the clear leader, driving around 77% of all AI traffic worldwide.

Other engines like Perplexity, Google’s Gemini, and Microsoft’s Copilot share the remaining 22% combined. Now, as of late-2025, we don’t have a first-party tool from OpenAI to see how people are searching for your brand (=prompt data). But we’re not completely blind. GA4 can still give you the important data on traffic from non-Google AI engines (traffic & engagement): You can use filters to see traffic from ChatGPT or Perplexity separately. You can compare it to your usual organic sources like Google or Bing. Break it down by location. And most useful for SEOs right now: seeing which specific pages are getting traffic from these AI engines and what the dynamics are. To get more details on your performance in ChatGPT or AI Mode and be able to compare your brand to your competitors, tools like SE Ranking or SE Visible provide data on AI visibility, mentions,

sources, and beyond. More on that in a moment. Before we get into specific tools, let’s talk about what metrics actually matter when you’re analyzing AI search visibility. If you want the most accurate data, start with traffic combined with conversions (or other business metrics, as well as engagement — broken down by channel or source as well as URL or country. Check how much traffic you’re getting from each channel and compare them — traditional search, AI engines — and how that traffic translated into your business performance. You can find all this in GA4 (and also visualize or merge this data using Looker Studio). That’s your hard data: how much traffic you get, how long people stay, how they interact, whether they convert, and how mu ch you earn from that traffic. - For organic traffic from Google, Search Console is the most accurate data source. Next, there is brand visibility, which is becoming

the go-to metric for AI search. While being rather probabilistic, it tells you how visible your brand is in AI-generated results — how often your brand is mentioned, whether your website links appear, what is the context around that (sentiment/the answers themselves) how you compare to your competitors and how that presence changes over time. And finally, there’s rankings for keywords/prompts— still useful, though less critical for AI search than for traditional SEO. Tracking your positions helps connect what you see in AI results with your overall search performance–but don’t rely too much on isolated positions. These last two metrics are what tools like SE Ranking currently help track — to see if your brand is part of the AI conversation. Now let’s get back to talking in detail about GA4. You can have different workflows to access

data around AI traffic in GA4. Here are some of the options. See your overall traffic Go to Acquisition → Traffic acquisition. Then click “+” to select → Session source or just change the first column to Session source. Here, you’ll see the broader pic1ture of how many sessions came from ChatGPT or other AI engines compared to your regular organic sources like Google or Bing or other channels. You can also add here the column by country or other information to get more insights. Want to check traffic by URL? Go to Reports → Engagement → Landing Page. This view shows which specific URLs users landed on after coming from AI engines. You’ll also find similar info in Pages and Screens. For better insights, I recommend setting the date range to around three months — AI traffic is still small, so you’ll get a more accurate picture with a wider window.

Next, filter your data: Choose Session source/medium → exactly matches → value=chatgpt|perplexity|claude|gemini|(other), and you’ll see all sessions per URL from AI tools only. You can also add filters by pages (e.g. to see ChatGPT traffic to your blog only) Finally, you can export the report: Click Share → Export to Google Sheets → press Import document. If you want to dig a little deeper, open the User or Tech tabs - they’ll show things like device type or audience data (like Age). Just keep in mind that demographic data isn’t always available for AI referrals because of limited tracking and privacy rules. You can also build a custom report in the Explore section — for example, to focus on traffic from AI engines only. You can create a separate report here with the breakdown of AI engines or URLs where traffic comes from AI engines only. And here’s another pro tip shared by Dana

DiTomaso that I really like–the way to get some traffic insights from AI Overviews in GA4. You can actually use GA4 to partly track traffic coming from AI Overviews by spotting URLs with fraggles. These are those long URLs with an additional piece of information , and they often appear in AIO sources. It won’t show you all the AI Overview traffic, but it’s a smart way to separate regular Google clicks from AI Overview visits. Then, you can build custom reports around them to better understand how AI Overviews might be impacting your traffic. If you want to see exactly how it works in practice, check out Dana’s breakdown - I’ll link the video in the description. And here’s the part we promised - SE Ranking AI Toolkit, designed to fill some of those data gaps we just talked about. It doesn’t give you “real traffic” numbers, but it gives you the next best thing: visibility data for any domain and any keywords or prompts. It shows your brand’s presence in AI search results,

for both mentions and links - and it’s a great way to benchmark your visibility, compare against competitors, and track your performance trends. Let’s break it down. First, SE Ranking’s AI Results Tracker. It does exactly what it sounds like — it monitors your presence in AI Overviews, AI Mode, ChatGPT or Gemini for your keywords. You’ll see: Mentions – how often your brand is cited in AI answers Link Presence – how often your URLs appear as clickable sources Positions of both mentions and links Presence in Top 3 – your share of top visible results You can also explore the AI answer for each day in detail and see lists of competitors appearing for your keywords. The feature I use most and really appreciate is the ability to export all links for all my tracked keywords. I used to do it manually to build a list of the most cited sources and brands. But there's now a separate Sources report that shows where your brand should be mentioned to get featured in AI answers more often.

Now, let’s see how to set up AI results tracking in SE Ranking. Step 1. Set up your project in SE Ranking if you haven’t already. Add your website, choose your search engines, and pick a region. Then, add your target keywords or prompts and competitors you want to monitor. Don’t worry about getting everything perfect at once - you can always update it later. Let it run for at least a month to get solid trends. Here, you’ll see two tabs: Classic Search and AI search. Add queries to all AI engines of interest–you can choose to track them across AI Overviews, AI Mode, ChatGPT, and more coming soon. In the case of AIOs, start with keywords that have informational intent and decent search volume. These are more likely to trigger AIOs and generate clicks, as our multiple studies confirm. Make sure to track keywords where you already have clicks from. If you’re exploring ChatGPT visibility, think of “keywords” more like prompts - since that’s

how users actually search in conversational AI chats. For AIO and AIM, you can track both but I still prefer keywords for now (but including long-tail questions like queries). The goal here is to see: Which AI results your keywords or prompts trigger Where your site or brand appears And how that presence changes over time If the Tracker gives you the keyword-level view, the AI Search Research tool zooms out to show the big picture. It’s perfect for brand-level and competitor-level analysis - giving you a high-level overview across AI Overviews, AI Mode, ChatGPT, Gemini, and Perplexity. For example, for AIO analysis you get the following: Mentions – total brand appearances in AI-generated text Link Presence – how often your site is cited AIO Traffic – estimated traffic potential from AI results Traffic Dynamics – trend of overall, organic, and AIO-sourced traffic Link Presence Dynamics – trend of link counts

in AIOs and their average positions over time This tool also shows all keywords where your site appears, including which URLs are cited, and all AI answers saved. You can also compare your domain with competitors’ to see where you’re strong, where competitors outperform you, and where new opportunities are emerging. So, while GA4 shows you real user traffic, SE Ranking’s AI Toolkit shows you the why behind it - how AI systems are referencing you, how often, and how that visibility evolves. And if you want to take this even further - there’s a really cool workaround. You can now connect SE Ranking’s API directly to Claude, ChatGPT, Gemini, or any other AI assistant through MCP. This gives you access to the same organic data on any domain (both from regular Google search and AI traffic), we’ve just been talking about - but pulled automatically, in real time. For instance, you can ask Claude,

in plain English, to pull live data for any domain - including metrics like AIO traffic, link frequency, average position within AI results, and more. It can even show month-over-month trends, helping you see how your AI visibility changes over time and how it compares to your organic performance - and you can keep analyzing it right there in chat. Moreover, if you want to automate the full analysis workflow, try connecting the API through services like make.com - so your visibility and AI traffic reports update themselves while you focus on strategy. Once you’ve collected your data from tools like GA4, GSC, or SE Ranking, it’s time to roll up your sleeves and dig in. Here's how I typically go about analyzing and enriching it—with help from AI, of course. I usually start by exporting data from GA4 (e.g., ChatGPT traffic) or Google Search Console (I also often use the Search Analytics for Sheets addon for getting this data filtered directly in Google Sheets).

From there, I merge it with regular rankings data by URL and add any other internal data (like page category or business metrics). To go even deeper, you can add more layers: Publication date to check how fresh the content is Status code checks (to catch 404s or incorrect redirects) Backlink data (to evaluate off-page impact) Etc. Pro tip: using data from GA4, you can easily find hallucinated pages from ChatGPT (and you should definitely do that as this AI engine quite often shows your non-existent URLs.) How to do that? Export all URLs with traffic, and combine this data with the status code (you can also add backlinks and internal links for additional check). Then, you’ll see where to set redirects if you have relevant pages or find new ideas for content. Alternatively, you can use Looker to merge the data or visualize it. You can connect data from GA4 or GSC and segment it there by page type or language

version by using custom dimensions, or merge this data with your internal data by using the blend feature (e.g. on the business metrics or page category). This way, you can quickly check organic data by a specific local version or page type (like your homepage, blog, or landing pages) to better understand what has grown or what specifically has dropped and make a decision based on that. Now, let’s move from tools that serve as data sources of AI traffic to AI tools that help you analyze the traffic data. ChatGPT is the next tool I make the best use of. For example, you can ask ChatGPT to write any Google Sheets formula you need—from something simple like counting words in a URL to more complex REGEX patterns for filtering pages or keywords. By linking the OpenAI API to your Google Sheets via Apps Script, you can write prompts directly in a cell and drag them down to apply across multiple rows (but now you also have the default =AI formula to do the same for you).

For example, you can use AI directly in Sheets to clean keyword lists or generate elements like titles or meta descriptions in bulk using your keywords. You can also do tasks like detecting a page's language version or page category directly in Google Sheets. But in cases like this, it’s usually safer to have ChatGPT write the formula and then paste it in yourself. You know how these things can be. And, you can also ask ChatGPT to build custom Google Sheets functions for even more automation. Let’s say you need to check the status code of a URL right inside Sheets. You don’t need to know JavaScript, instead just ask ChatGPT to write the script for you. Try a prompt like: “Write a Google Sheets App Script function that checks the HTTP status code of a URL and returns it in the given cell.” If the script doesn’t work right away, just paste the error back into ChatGPT—it’ll help you debug it.

Once your script is ready: Open your Google Sheet and go to Extensions → Apps Script. Paste the code into the editor. Add your API key and deploy it. Then, use the new function like any formula in your sheet. Reference the cell with the URL, and drag down for bulk actions. Done! Claude is another option that works well for more complex data analysis. It can take your GA4, GSC data exports, as well as data from SE Ranking or other SEO tools and turn them into clear summaries or charts with key insights. To get started: Add a clear prompt (e.g., “Summarize this dataset of three months of traffic and registrations by URL and give me the comparison of MoM performance, for the homepage, blog, and non-blog separately.”) Export your data as a CSV and upload the file to Claude with your prompts Human may now rest. Task delegated! Here’s what you can also ask Claude to do with your exported data:

Calculate basic statistics, like mean, median, minimum, and maximum for both clicks and impressions. Identify top pages like your top 10 URLs by clicks and your top 10 by impressions or top growing pages QoQ or MoM. Group and categorize URLs into types such as blog posts, product pages, or category pages. You can also ask it to label them as high, medium, or low performers based on click or impression thresholds. Merge and compare data from different search engines Visualize your data–even using your brand colors One quick mental note: Before you let it crunch the numbers, do a quick sanity check. Ask for a quick summary: total number of URLs, total clicks, the usual. That way, you can make sure it’s reading the file correctly before diving deeper. And that’s it for this part, hope you found it helpful! But there’s plenty more expert-level insight waiting for you in our AI for SEO course at the SE Ranking Academy. Go check it out!