To audit your brand's presence on ChatGPT, Gemini, and Perplexity, run 10 to 15 industry questions across all three platforms in a private browser, repeat each question three times to account for variance, and score every answer on a zero-to-five rubric based on whether your brand is mentioned, cited, or recommended. Log the competitors and sources named alongside you. The full process takes about 90 minutes, uses only free tiers, and produces a baseline visibility number you can track quarterly.
That is the short version. What follows is the full framework, with the specific questions to ask, how to interpret the answers, and what to do with the results.
Want to get citied on AI Platforms, Consult PPC Guru Now!
**Key takeaways **
- ChatGPT reached 900 million weekly active users in February 2026. Google Gemini's app crossed 900 million monthly active users at Google I/O in May 2026, and Google AI Mode inside Search now serves over one billion monthly users.
- About 37% of consumers now begin their searches with an AI tool rather than a traditional search engine
- Google AI Overviews appear on roughly 48% of tracked queries as of March 2026, climbing to 68% for local business queries and 88% for healthcare
- The same question asked twice on the same AI platform can produce different brands, different citations, and a different answer. That is why manual audits with repeat runs beat single-shot automation.
- Only 14% of marketers currently track AI search performance. The first-mover window is still open.
Why AI search visibility matters in 2026#
AI search is where a growing share of buying decisions now start. ChatGPT alone processes over 2 billion queries per day. Perplexity, though smaller at roughly 30 million monthly active users, is the fastest-growing specialist search engine on the web.
For local service businesses, the exposure is larger than most owners realize. According to research, 540-query local search study found AI Overviews on 68% of local business queries on average, jumping to 92% for informational queries like "how long does an eye exam take near me" and 97% for hybrid queries like "average cost of dental implants in Phoenix." Restaurant queries trigger AI Overviews 78% of the time. Healthcare hits 88%. B2B tech hits 82%.
There is a wider structural point. The overlap between Google's top-10 organic results and AI citations dropped from about 75% in mid-2025 to somewhere between 17% and 38% in early 2026. Ranking on page one no longer means you get pulled into the AI answer above it. Different game, different signals.
What is an AI brand visibility audit, and why do it manually?#
An AI brand visibility audit is a structured check of whether large language models mention, cite, or recommend your business when someone asks a question your customers ask. Manual means you type the questions yourself and read the answers, rather than running everything through a tracking tool.
Manual audits matter because AI answers are probabilistic. The Princeton and Georgia Tech KDD 2024 study on generative engine optimization confirmed what most practitioners see day to day: the same prompt returns a different distribution of brands and citations across runs. Check once, and you have photographed a single roll of the dice.
There is a second reason. Running the audit yourself teaches you where the models are actually getting their information. Muck Rack's December 2025 study found that 82% of AI citations come from earned media rather than owned content. Watch a model cite a Reddit thread ahead of your carefully written service page twice, and you will restructure your entire content plan.
How to set up your audit environment before starting#
Do this once, then you can rerun the audit any time in under an hour. Open a private or incognito browser window. Log out of ChatGPT, Gemini, and Perplexity, or use a browser you never sign in with. Personalization skews the audit, because platforms remember what you have asked before and adjust responses toward you.
Set your location Perplexity uses IP-based location signals, and Google AI Overviews weight Google Business Profile data heavily. If you serve Toronto and audit from a VPN pointed at Chicago, the results are worthless.
Build a simple spreadsheet with these columns: question, platform, run number (1, 2, 3), brand mentioned (yes or no), competitor names, cited sources, and score (0 to 5). Nothing fancy. Google Sheets works.
Step 1. Build a 15-question prompt list across four categories#
Do not invent questions in your head. Write down what customers ask you in first meetings and translate those into search language. Cover four categories with two to four questions each.
Discovery questions are how prospects start when they do not yet know who to hire. Examples: "Best digital marketing agency in Toronto." "Top HVAC companies in Mississauga." "Immigration consultants near me." These are the queries where you either exist in the answer or you do not.
Comparison questions are what buyers ask when they are down to a short list. "Google Ads agency vs Meta Ads agency for a small business." "Full-service marketing agency vs freelancer for a startup." Comparison prompts produce the richest citations because the model has to work harder to give a balanced answer.
Problem-first questions are the highest-intent category, and most brands ignore them. "Why is my Google Ads cost per lead going up." "How do I get more customers as a plumber in Toronto." "What should I do if my Meta ads are getting clicks but no leads." When you appear in a problem-first answer, you are being handed to somebody who already knows they need help.
Trust and reputation questions are what happens after somebody has heard your name. "Is PPC Guru a legitimate marketing agency." "Reviews of Acme HVAC Toronto." These show what the model has learned about you from third-party sources.
Stop at 15. Beyond that, you are testing yourself, not your visibility.
Step 2. Run each prompt three times on each platform#
Run every question three times per platform, in fresh sessions where possible. Cover ChatGPT, Gemini, and Perplexity at minimum. Add Copilot if your customers work in Microsoft-heavy environments, and Claude if they are in professional services or tech.
Three runs surface three patterns. Some brands appear in every run. Those brands own the topic. Some appear in one run out of three. That is a coin flip, and it means the model has weak associations for that query. Some brands never appear at all.
Run once, and you will conclude either that you are visible or that you are not. Neither is true. What is true is a probability, and probabilities are what you plan against.
Step 3. Score every answer on a zero-to-five rubric#
Heading#
Use the same rubric every time. Consistency is what turns a subjective read into a trackable number. 0: brand not mentioned at all 1: brand mentioned only after a direct follow-up question 2: brand mentioned in a list, not first, no citation 3: brand mentioned in a list, with a citation to your site 4: brand mentioned in the main body of the answer, with a citation 5: brand is the primary recommendation, cited, and the facts about your business are correct Average the score across the three runs per question, then across all 15 questions. That number is your baseline. If it moves 0.3 points over a quarter, that is real movement. Track it.
Step 4. Track competitors and cited sources alongside your score#
Heading#
While you score your own visibility, log every competitor named in every answer and every source cited. This is often the most useful output of the whole exercise.
The competitor list tells you who the models think you are up against. Sometimes those are the businesses you already know. Sometimes a directory listing or a national chain you had dismissed is being surfaced ahead of you. The citation list tells you where the models learned it from. Often it is not a competitor's website at all. Reddit threads, industry publications, Google Business Profile pages, Trustpilot, and G2 profiles show up constantly.
Semrush's 2025 citation analysis found that the top 5 cited domains capture 38% of all AI citations, and the top 20 capture 66%. A small number of sources are doing most of the work. If you are not among them and your competitors are, you have just identified your next quarter's content and PR targets in about two minutes.
Step 5. Turn the audit into three decisions#
Heading#
An audit that does not lead to action is a report. Force three decisions from what you found.
Decision one: content. Pick the three questions where you scored lowest and where buyer intent is highest. Those are your content priorities. Write answers to them with the direct answer sitting in the first two sentences under each heading. That format is what models pull citations from. Princeton's KDD 2024 study measured that inline citations lift AI citation likelihood by around 30%, and quotation addition by 41%.
Decision two: earned media. Pick two cited sources where competitors appear and you do not. Pitch a guest article, claim a directory listing, or ask a happy client to leave an honest review on a platform the models read. Muck Rack found that third-party trust signals raise AI citation likelihood roughly 75x. This is the single highest-return move in the whole audit.
Decision three fix wrong facts. If any answer said something inaccurate about your business, correct the source. AI Overviews draw heavily from Google Business Profile data, structured data on your own site, and third-party citations. Fix the profile, add FAQ schema to the relevant service page, update the outdated bio wherever it lives.
How often should you re-run the audit?#
Heading#
Every 90 days is a reasonable cadence for most small businesses, and matches how fast the underlying models change. Ahrefs' 17-million-citation study in early 2026 found that 76.4% of top ChatGPT citations come from content updated within the previous 30 days. Content freshness is not a suggestion. If you audit once and forget, you will watch your baseline erode without knowing why.
Frequently asked questions#
Heading#
Do I need a paid subscription to audit ChatGPT, Gemini, and Perplexity? No. Free tiers are what your customers use, and paid personalized accounts distort the audit. Run everything from an incognito window.
How is this different from a keyword ranking report? Keyword rankings measure a deterministic system. AI answers are probabilistic. You are measuring a probability distribution across brands and citations, which is why running each prompt three times matters.
What if my industry does not trigger AI Overviews often? Some categories still show them rarely. Simple local-intent queries like "plumber near me" trigger AI Overviews only about 7% of the time. The standalone AI tools your buyers use during research do not care about local pack rules. Audit them anyway.
Which platform matters most? It depends on your buyer. ChatGPT leads U.S. AI chatbot market share at roughly 58% as of May 2026, followed by Gemini at 19% and Claude at 13%. Perplexity captures a smaller slice but converts higher, around 10.5% compared with Google organic's 1.76%. If you serve high-consideration buyers, Perplexity matters more than its user count suggests.
Closing thought#
The businesses winning inside AI answers today are not the biggest. They are the ones that looked, honestly, at what the machines are saying about their category, then made a short list of changes based on what they saw. Ninety minutes with a browser and a spreadsheet gets you to that list. Everything after that is execution.
Author bio: Siddharth Sharma is an SEO Specialist at PPC Guru, a Toronto-based digital marketing agency offering SEO, PPC, Google Ads, and social media marketing services. With 6+ years of experience in search and digital marketing, he focuses on helping small and mid-sized businesses build sustainable organic visibility. He writes about SEO, AI search, and practical digital marketing strategy.