Measure AI visibility one platform at a time, because each AI platform shows your business differently and gives you different data. Google reports AI Overviews and AI Mode impressions in Search Console. Bing reports Copilot citations in Bing Webmaster Tools. ChatGPT, Gemini, Claude and Perplexity show up as referral visits in Google Analytics 4. For everything the platforms do not report, such as whether an AI tool names your business at all, you need a repeatable sample of prompts.
Start with one question, because it decides where your measurement time goes: which AI platform do your customers actually use before they contact you? Ask your last 20 enquiries. A Coimbatore manufacturer whose buyers use ChatGPT needs a different set-up from a clinic whose patients see AI Overviews on Google.
Technox offers SEO & AI SEO services, so we have an interest in this topic. We do not sell an AI visibility tool, and most of what follows can be done with free tools and a spreadsheet.
What AI visibility means, and why one score is not enough
AI visibility is how often, how prominently and how accurately AI tools present your business when people ask questions you could answer. If you are new to how these systems choose sources, start with how AI search works.
The IAB, the digital advertising industry body, published a measurement framework in August 2026 that splits AI visibility into four parts, which it calls the 4 P's (AdExchanger, August 2026):
Presence: how often your brand appears, or your content is cited.
Prominence: where it appears, and whether it is highlighted or listed alongside rivals.
Portrayal: how you are described, including sentiment and factual accuracy.
Persuasion: whether the mention sends visits, for example through post-citation click-through rate.
The IAB also notes that more than 20 companies now sell AI visibility measurement tools, using methods that can produce different answers for the same brand (IAB, August 2026). That is the first reason not to trust a single blended score.
The second reason is that the platforms do not agree with each other. Profound analysed 100,000 prompts and found that only 11% of the domains cited appeared in both ChatGPT and Perplexity answers. Overlap between other pairs ranged from 6% (Google AI Overviews and Microsoft Copilot) to 16.4% (Perplexity and AI Overviews) (Profound, July 2025). Being visible on one platform tells you little about the others.
The third reason is that answers change from one run to the next. In a study SparkToro and Gumshoe published in January 2026, 600 volunteers ran 12 prompts nearly 3,000 times across ChatGPT, Claude and Google's AI. The odds of getting the same list of brands twice were under 1 in 100, and the odds of the same list in the same order were closer to 1 in 1,000 (Search Engine Land, 2026). The study also found that smaller markets, such as regional service providers, gave more stable results. That is useful news for a business competing in Coimbatore rather than across India.
Why measure platform by platform
Each platform gives site owners a different slice of data, and none gives the full picture. The table below is the map the rest of this guide follows.
The audiences behind these platforms are large. Google says AI Mode passed 1 billion monthly active users (Google, July 2026), and the Gemini app reached 1 billion monthly active users by October 2026 (Google, October 2026). OpenAI said in February 2026 that ChatGPT had 900 million weekly active users (TechCrunch, February 2026).
The clicks are still small. Conductor's 2026 benchmarks report, covering 13,770 US domains and more than 3.3 billion sessions from May to September 2025, found AI referrals made up 1.08% of website traffic, and ChatGPT accounted for 87.4% of those AI referrals (Search Engine Land). Big audience, few clicks: that gap is why you measure visibility and visits separately.
For Indian businesses, two facts matter. Google added Tamil, along with Bengali, Kannada, Malayalam, Marathi, Telugu and Urdu, to AI Mode in October 2025 (Google India, October 2025), so your customers may be asking in Tamil. And Airtel offered a free 12-month Perplexity Pro subscription to all of its 360 million customers in July 2025 (Bharti Airtel, July 2025). Do not assume your customers only use ChatGPT.
The framework in five steps
The framework works in this order. Each step uses free data first and only adds paid tools where the free data runs out.
For step 1, add an "AI assistant" option to the "How did you hear about us?" question on your enquiry form, and ask the same question on calls and WhatsApp enquiries. Name the tools in the options (ChatGPT, Google's AI answer, Gemini, Perplexity, Copilot), because customers may not think to name the platform unprompted.
Steps 2 to 4 are covered platform by platform below. Step 5 has its own section.
How to measure Google AI Overviews and AI Mode
Use the Generative AI performance report in Google Search Console. It covers AI Overviews and AI Mode and shows impressions, which Google defines as how many times links to your site were shown to a user in a generative AI feature on Google Search (Search Console Help).
What you need to know about this report:
Google introduced it in June 2026 and had released it to sites globally by September 2026 (Search Engine Roundtable, September 2026).
It shows impressions only. There is no click data and no query data.
You can break impressions down by page, country, device and date, and filter text searches from multimodal ones such as Google Lens.
Several links from your site in one AI feature count as one impression in the chart, so chart and page-level totals can differ.
The data is also included in the Web search type of the normal Performance report, not separate from it.
The report may not appear if your site has too few AI impressions, and it shows nothing if your site is excluded from Google's generative AI features.
Clicks from AI Overviews and AI Mode do not show up as a separate channel in GA4. They are counted as Organic Search, not under GA4's new AI Assistant channel (Search Engine Journal, July 2026).
How to use it each month:
Export AI impressions by page and compare them with each page's total Web impressions.
List the pages that get AI impressions and the pages that get none. The difference shows where Google already uses your pages in AI answers and where it does not.
Filter by country if you serve more than one market. For Technox, that means India and Malaysia separately.
For queries, use your prompt sample (step 3), because Search Console will not tell you which questions triggered the AI answer.
For what to change once you have the data, see our guide to Google AI Mode optimisation.
How to measure Microsoft Copilot and Bing
Use the AI Performance report in Bing Webmaster Tools. Bing launched it as a public preview in February 2026, and it covers Microsoft Copilot, AI-generated summaries in Bing and select partner integrations (Bing Webmaster blog, February 2026).
It reports four things:
Total citations: how many times your pages were shown as sources in AI answers.
Average cited pages: how many unique pages from your site were cited per day.
Grounding queries: a sample of the phrases the AI used when it retrieved your content.
Page-level citations: citation counts for each URL.
Bing is clear that these counts do not show where or how a citation appears in the answer, or how important a page is.
In October 2026, Bing added Topics Portfolio, a beta view that groups grounding queries into topics and shows Opportunity Queries, where other sites are cited and yours is not (Bing Webmaster blog, October 2026). It is the only first-party competitor view we found on any AI platform. Bing itself says to treat the numbers "as directional signals for planning, not as a precise ledger."
How to measure ChatGPT
ChatGPT does not give site owners a dashboard, so you measure it through visits, logs and prompts.
Visits in GA4. In May 2026 Google Analytics added an AI Assistant channel to the default channel group. Recognised AI referrals get the medium "ai-assistant", and Google named ChatGPT, Gemini and Claude as examples, without publishing the full list (Search Engine Journal, May 2026). Visits with no referrer, common from mobile apps and in-app browsers, still land in Direct.
The utm_source tag. ChatGPT normally adds utm_source=chatgpt.com to the links it sends, which lets GA4 identify the visit even without a referrer. Check your reports for the week of 8 to 15 October 2026. After GPT-6 rolled out, many ChatGPT links opened without the tag or a referrer, and Graphite's Ethan Smith reported falls of 22% to 57% in ChatGPT sessions for sites with meaningful AI Assistant traffic, starting 9 October. Adobe's Cedric Huesler said OpenAI told him privately it was a bug and a fix began on 15 October; OpenAI had not commented publicly at the time (Search Engine Journal, October 2026). Add an annotation for that week so nobody reads it as lost visibility.
Server logs. OpenAI documents three agents worth separating in your logs: OAI-SearchBot, which surfaces websites in ChatGPT's search features; GPTBot, which OpenAI uses to improve its foundation models; and ChatGPT-User, which visits pages for certain user actions in ChatGPT and custom GPTs. OAI-SearchBot visits tell you whether ChatGPT search can reach your pages. ChatGPT-User visits tell you your pages are being fetched for user requests. Our post on technical SEO for AI search visibility covers the robots.txt side.
Mentions without a link. ChatGPT can name a business without linking to it. Only a prompt sample will catch that. For how to earn those mentions, read how to rank on ChatGPT.
How to measure Perplexity, Gemini and Claude
These three are measured the same way as ChatGPT: referral visits, crawler visits and prompts.
Perplexity. Perplexity's referrals may not be covered by GA4's AI Assistant channel; one analysis found it missing from Google's recognised list as of June 2026 (Search Engine Journal, July 2026). In logs, look for PerplexityBot, which surfaces and links sites in Perplexity results, and Perplexity-User, which fetches pages when a user asks a question and generally ignores robots.txt rules because the user requested the fetch (Perplexity docs).
Gemini. Gemini app referrals come from gemini.google.com, and Google names Gemini as one of the sources its AI Assistant channel recognises. Logs help less here. Google-Extended, the robots.txt control for Gemini, has no separate user agent, because Google crawls with its existing user agents, and it does not affect inclusion or ranking in Google Search (Search Engine Roundtable, February 2024).
Claude. Anthropic runs three bots: ClaudeBot for training data, Claude-SearchBot for indexing to improve search results, and Claude-User when a person asks Claude a question that needs a web page (Anthropic, updated April 2026). Referrals come from claude.ai. Google named Claude when it launched the AI Assistant channel, but one analysis reported it was no longer on the recognised list by June 2026, so include it in your own channel too.
To catch all of these in one place, create a custom channel group in GA4 (copying the default group is the quickest start), add a channel called "AI referrals", place it above Referral, and set its condition to session source matching this regular expression:
chatgpt\.com|openai\.com|perplexity|gemini\.google\.com|copilot\.microsoft\.com|claude\.ai|deepseek|grok
Review the list every quarter, because new assistants appear and referrer domains change.
How to build a prompt sample that means something
A prompt sample is a fixed list of questions you put to each AI platform every month, recording what comes back. It is the only way to measure presence, prominence and portrayal on platforms that report nothing.
Size matters. The IAB classes fewer than 50 queries as "exploratory" and says that is too few to characterise a category; only larger, tested samples across more platforms count as "decision-grade" for budget decisions (AdExchanger, August 2026). Because answers vary between runs, the SparkToro study's advice is to track how often your brand appears across many prompts run many times, not your position in one answer.
Here is how we build one for a local business:
Write at least 50 prompts and split them by buying stage. Early-stage prompts describe a problem ("why is my website not getting enquiries"). Comparison prompts name the service and place ("best web development company in Coimbatore"). Brand prompts name you ("is Technox good for Shopify development").
Add local variants. Use areas customers actually search from, such as Gandhipuram, RS Puram, Peelamedu or Saravanampatti, and include Tamil and Tanglish phrasings if your customers use them. If you serve another market, as Technox does in Kuala Lumpur, write a separate set.
Run each prompt on each platform that matters to you, three to five times a month, from a logged-out browser where the platform allows it. Note the location setting each time.
Record the same fields every run, using the log below.
Report mention rate as answers that mention you divided by total answers, and always show the total. "Mentioned in 41 of 250 answers" is honest; "16% visibility" on its own hides the sample size.
Do not skip the accuracy column. An AI answer that lists a service you stopped offering, or an old address, sends customers away before they reach your site. Our post on entity SEO explains how AI tools build their picture of your business.
Paid tools such as Semrush's AI Visibility Toolkit, Ahrefs Brand Radar, Profound or Peec AI automate the running and recording. They are worth it once you track hundreds of prompts or several competitors. Check each tool's method before you trust its score, for the reasons the IAB gives above.
How to connect AI visibility to leads
AI visibility only matters if it brings enquiries. Wil Reynolds of Seer Interactive argues that AI visibility on its own is a vanity metric, and suggests pairing it with branded search, direct traffic and the pages AI visitors land on (Seer Interactive, February 2026). In one Seer study of a B2B client's customers, 44% of real prompts included brand names, which Seer reads as people already trusting a brand before they ask AI about it.
Four signals connect visibility to business:
Self-reported source: the "How did you hear about us?" answers from step 1. This is the most direct evidence you will get.
Branded search: filter Search Console queries by your brand name and watch impressions over months. People who meet you in an AI answer may search for you by name later.
AI landing pages: in GA4, check which pages AI referral visitors land on. Service and contact pages suggest buying intent; blog posts suggest research.
Conversions by channel: set up enquiry form submissions, calls and WhatsApp clicks as key events, then compare conversion rates for AI referrals against organic search.
A monthly AI visibility scorecard
Put the numbers from every step on one page and review it monthly. Look at trends over at least three months before changing course, because single months move around.
Common mistakes when measuring AI visibility
Trusting one score from one tool. Different vendors' methods give different answers for the same brand.
Tracking rank position. Answers change between runs, so how often you appear is more reliable than where.
Reading a tracking change as a visibility change. The October 2026 ChatGPT tagging bug is the latest example.
Expecting AI Overviews clicks in the AI Assistant channel. GA4 counts them as Organic Search.
Blocking the bots you want to measure. A robots.txt rule meant for training crawlers can also block search crawlers if it is written too broadly.
Measuring with ten prompts. Below 50, you are exploring, not measuring.
What to do after you measure
The scorecard tells you which problem you have. Each one has a different fix:
Not mentioned anywhere: start with the basics in what is AI SEO and the SEO and AI visibility checklist.
Mentioned, but described wrongly: fix the sources AI tools learn from. Our post on whether brand authority matters for AI visibility covers this.
Cited, but no visits: improve the cited pages so they answer the question and give a reason to click. See on-page AEO.
Crawlers not visiting: check robots.txt, firewall rules and rendering, as covered in technical SEO for AI search visibility.
Online store: product data has its own issues, covered in AI SEO for ecommerce.
If you are still deciding whether this is a separate job from SEO, our comparison of AEO vs SEO explains where they overlap.
Where Technox fits
You can run this framework yourself. Search Console, Bing Webmaster Tools and GA4 are free, and a prompt sample needs a spreadsheet and a few hours a month.
Technox can help in three places: setting up the GA4 channels, events and Search Console exports so the data is reliable; building and running a prompt sample for Coimbatore, wider India or Malaysia; and fixing what the scorecard shows. If you want a second opinion first, ask for a free audit or contact us. Our AI SEO services in Coimbatore page explains how we work.
Frequently asked questions
What is AI visibility?
AI visibility is how often, how prominently and how accurately AI tools such as Google AI Overviews, AI Mode, ChatGPT, Gemini, Perplexity, Copilot and Claude present your business when people ask relevant questions. It is measured through platform data, referral visits and a repeatable sample of prompts.
Can I see AI Overviews traffic in Google Search Console?
You can see impressions, not clicks. Search Console's Generative AI performance report shows how many times links to your site appeared in AI Overviews and AI Mode, by page, country, device and date. It does not show clicks or the queries that triggered them.
Does Google Analytics 4 show ChatGPT traffic?
Yes. Since May 2026, GA4's default channel group includes an AI Assistant channel for recognised AI referrals; Google named ChatGPT, Gemini and Claude as examples at launch. Visits that arrive without a referrer or tag still appear as Direct, and a custom channel group catches sources the default list misses.
Why did my ChatGPT traffic drop in October 2026?
Probably because of tracking, not lost visibility. After GPT-6 rolled out on 7 and 8 October 2026, many ChatGPT links opened without the utm_source tag or a referrer, so visits showed as Direct. A fix was reported to have started on 15 October. Annotate that week in your reports.
How many prompts do I need to track AI visibility?
At least 50. The IAB treats fewer than 50 queries as exploratory and too few to characterise a category. Run each prompt several times a month on each platform you care about, because AI answers change from one run to the next.
Are AI visibility tools accurate?
They are useful but not definitive. The IAB notes that more than 20 vendors sell AI visibility tools and their methods can give different answers for the same brand. Check how a tool samples prompts and repeats runs before you rely on its score.
Is there a free way to measure AI visibility?
Yes. Google Search Console, Bing Webmaster Tools and Google Analytics 4 are free, and a monthly prompt sample needs only a spreadsheet. Paid tools become worthwhile when you track hundreds of prompts or several competitors.
How often should I measure AI visibility?
Monthly for platform data, referrals and your prompt sample, and quarterly for server logs. Judge trends over at least three months, because single months vary.




