How to track LLM visibility in ChatGPT, Gemini and Claude, and analyze your competitors' visibility over time

The case for why AI search visibility matters more than Google rankings is already on the blog. This post is the procedure. Your Google rank takes a minute to check. Your ChatGPT rank has no results page at all. A shopper asks for a cast-iron pan that works on induction and gets one written answer naming a handful of brands. You are in it or you are not.

Adobe reported in January 2026 that over the holiday season, "traffic to retail sites from generative AI tools (shoppers clicking on a link to a retail site) increased by 693.4% compared to the year prior." Shoppers are asking. Can you see what they are told?

How to measure AI search visibility inside an AI assistant

A search engine returns a list and you hold a slot on it. An assistant writes one answer, so the unit of measurement is the answer: were you in it, where, in what tone, and next to whom. That is AI visibility in practice.

Bain's April 2026 advice to brands is to measure generative engine performance across top personas, categories and prompts, tracking share of voice, citation frequency and sentiment trajectory across engines. In sheet terms that is a mention column for share of voice, a cited-sources column for citation frequency and a sentiment column, per prompt per engine. This guide adds position and competitors named, because those are what show an overtake.

Step 1: build the prompt set from real buying questions

Rewrite each category keyword you rank for or bid on as a sentence a shopper would type into a chat box. Shopify's advice: "You can convert your existing list of keywords into a set of prompts, or you can do some searching in an AI answer engine yourself." Do both. Typing into an assistant turns up phrasing you would not guess.

Adobe's survey of 5,000 U.S. consumers, published in March 2025, listed the shopping tasks people use generative AI for: "conducting research (per 55 percent of respondents), receiving product recommendations (47 percent), seeking deals (43 percent), getting present ideas (35 percent)." So your templates are "best X for Y", "X vs Y", "is Brand worth it", "gift for someone who" and "X under [budget]".

Keep the set small enough to run by hand and wide enough to cover every category you make money in. Columns: prompt, engine, date, then one per signal. Freeze the wording; better wording goes in a new row.

Step 2: run ChatGPT visibility tracking alongside Gemini, Claude and Copilot

Run all four, because they answer from different places.

OpenAI asks merchants to supply product data to ChatGPT. Its commerce documentation says: "Provide a structured product feed so ChatGPT accurately indexes and displays your products with up-to-date price and availability." A stale feed is a ChatGPT problem that checking Gemini will never show you.

Claude searches the web when it judges the question needs it. In Anthropic's words, it then "provides direct citations so you can easily fact check sources." Read them when a competitor appears, to see which pages the answer drew on.

Then the discipline. Same wording every time and the same account state. If an engine offers a shopping mode or web-search toggle, pick one setting and keep it. Change the method and you break the trend.

Step 3: record four signals per run (mention, position, sentiment, competitors named)

Score by rules a new hire could repeat. Mention is yes if your brand is named anywhere in the answer. A link without your name does not count, because shoppers read names.

Position is your order in the recommended list, counting from one. If the answer is a paragraph, count the order in which brands are named. Leave it blank when you are absent; blank and zero mixed together make averages lie.

Sentiment is positive, neutral or negative, decided by the words around your name. "Reliable but pricey" is neutral. "Often recommended for durability" is positive. "Some reviewers report sizing issues" is negative. Paste the deciding phrase into a notes cell.

Competitors named is every other brand in the answer, in order. This column is where an overtake first shows. Add an optional fifth column for cited sources, the URLs the answer links to. Trackers label them differently, for example browsing sources or citation sources.

Step 4: set a cadence to track LLM visibility and ChatGPT rankings over time

Weekly is the honest cadence for a manual sheet: frequent enough to catch a competitor's content push, slow enough that one person keeps it up.

Read a four-run trend and treat a one-week move as noise until the next run confirms it. After a real event, such as a price change, re-run and note the event in an extra row.

If you automate, keep the discipline. AI Visibility Monitoring in RivalSweeper runs each prompt set daily, every three days, weekly or monthly, and on-demand runs feed the same history.

Competitive analysis of AI search visibility: computing share of voice

A competitive analysis of AI search visibility rests on two numbers. Mention rate is the prompts where you appear divided by the prompts run. Competitive share of voice is your mentions divided by all mentions across the brands you track. Shopify gives the general form, "Share of voice (%) = Your brand metrics ÷ Total market metrics × 100", and calls the LLM version a newer metric that measures your brand against competitors across AI platforms.

The table runs the arithmetic on made-up numbers. It is hypothetical, with no brand, client or engine data behind it.

Compute both per engine and blended. The blend goes in the monthly deck; the per-engine number tells you where to work, because a blend hides an absence on one engine.

How to track competitor rankings in LLMs and spot an overtake

To track competitor rankings in LLMs, wait for four or more runs, then watch for three patterns. The first is your mention rate flat while a competitor's climbs on the same prompts. Check their cited sources first for a new comparison page or a review roundup that now lists them.

The second is your position slipping from first to third while mention rate holds. You are still in the answer, but the framing moved. Look in the sentiment column for a new caveat next to your name.

The third is sentiment turning negative on one engine only. Look for a source that engine reads and the others do not, such as a forum thread. Reddit brand monitoring is the natural pair: when AI Visibility Monitoring flags a sentiment drop in an engine's response, the Reddit report usually shows the thread behind it. AI Visibility Monitoring also alerts when visibility drops or a competitor takes a slot you held.

Treat each pattern as a competitive gap analysis job: which categories they own in the answers, which you own, and where the whitespace is.

Why Google AI Overviews is a different surface from a model answer

The difference between LLM monitoring and AI search monitoring comes down to the surface. AI Overviews and AI Mode are part of Google Search. Google says sites appearing in AI features such as AI Overviews and AI Mode are included in overall Search Console traffic and reported in the Performance report under the 'Web' search type. Microsoft has an AI Performance report for Copilot in Bing Webmaster Tools in public preview. For what ChatGPT and Claude say about your category, a prompt sheet is the practical way to see the answers themselves. For the Google side, read how AI Overviews change ecommerce competitive intelligence, and use the client-ready explainer on AI Overviews vs AI visibility to explain the split to a client.

What to do when your brand is invisible

Diagnose by spread. Missing on every engine? Check for a data or access problem first. Missing on one? Check the sources that engine cites.

For the all-engines case, start with the six signals behind the Agent Readiness Score (0-100). It is calculated for every monitored domain, so add the competitor who keeps appearing in the answers as a monitored domain and compare the two scores. Run the free Agent Readiness check, and read the Agent Readiness Score deep dive for each signal. For ChatGPT, also check the product feed OpenAI asks merchants to provide.

For the one-engine case, pull the cited-source column for the prompts where you are missing and list the review sites and comparison pages you do not own. Those are the pages the engine reads. Which page types AI bots fetch covers what to publish on your own domain, and the generative engine optimization entry covers the discipline.

Doing this with a ChatGPT and LLM visibility tracker instead of a spreadsheet

The spreadsheet works until its owner gets busy and the trend dies. ChatGPT visibility tools compared lists the options by name.

In AI Visibility Monitoring you select the domain, add prompts, and RivalSweeper runs the same prompts against several large language models, including Google Gemini and OpenAI GPT models, on your schedule or on demand. Each run records share of voice, position, citation sources and sentiment per model, rolled into a 0-100 LLM visibility score. It does not query the ChatGPT, Claude or Copilot apps themselves, so checking those assistants directly stays a manual step: keep a short version of your sheet for them. The Competitor Comparison report shows share-of-voice deltas, sentiment deltas and citation-source overlap on the same prompts. The ChatGPT visibility tracker and Gemini visibility tracker pages list each engine's metrics. To go further, book a 15-minute demo or see plans.

Frequently asked questions

How can I analyze my competitors' visibility in LLMs?

Run your customers' buying prompts on ChatGPT, Gemini, Claude and Copilot and log every brand named, in order, with its sentiment. Their mention rate is appearances over prompts run; their share of voice is their mentions over all tracked brand mentions.

How can companies track their visibility inside AI assistants like ChatGPT or Gemini?

Keep a fixed prompt set and method: identical wording and account state, a logged date, and the same four signals per prompt per engine. Read trends over four or more weekly runs, and move to a scheduled tool when the sheet stops getting filled in.

How do I track my visibility and rankings in ChatGPT over time?

Treat "ranking" as your position in ChatGPT's recommended list for one prompt, blank when you are absent. Record it weekly the same way and chart position and mention rate across runs. Log cited sources too, so a drop can be traced to a feed or page.

How can I track my competitors' visibility in ChatGPT shopping recommendations?

Write category prompts in shopping form ("best X under a budget"), run them in ChatGPT on a fixed schedule, and record every competitor named, its position and the sources cited. If a competitor appears with current prices while you are absent, check your product feed first.

How do I monitor brand presence across ChatGPT, Claude, and Gemini?

Use one prompt set across all three and record the four signals per prompt per engine in one sheet. Compute mention rate and share of voice per engine before blending, so a collapse on one engine stays visible.

What is the difference between LLM monitoring and AI search monitoring for brands?

In this guide, LLM monitoring means checking the written answers of assistants such as ChatGPT, Claude, Gemini and Copilot with a fixed prompt set. AI search monitoring here means Google's AI Overviews and AI Mode, which Google reports inside Search Console. Microsoft also has an AI Performance report for Copilot in Bing Webmaster Tools, in public preview.

How do I track my brand visibility in LLMs without a tool?

Rewrite category keywords as shopper questions, run them weekly on ChatGPT, Gemini, Claude and Copilot with identical wording, and log the four signals in a spreadsheet. That is how to measure AI search visibility by hand.

References

• Adobe, holiday shopping season results, January 7, 2026: news.adobe.com
• Adobe Analytics, traffic to US retail websites from generative AI sources, March 17, 2025: blog.adobe.com
• Bain and Company, Your Next Customer Will Find You Using AI. Now What?, April 7, 2026: bain.com
• Shopify, Share of Voice Explained, September 11, 2025: shopify.com
• Shopify, AEO for eCommerce, February 13, 2026: shopify.com
• Anthropic, Claude can now search the web, March 20, 2025: claude.com
• OpenAI, commerce developer documentation: developers.openai.com
• Google Search Central, AI features and your website: developers.google.com