Back to BlogSEO and Digital Marketing

AI Visibility Audit: Your Brand in ChatGPT, Perplexity and Gemini

Jay PipaliyaPublished October 9, 202610 min read✓ Last Updated: 2026-10-08
AI Visibility Audit: Your Brand in ChatGPT, Perplexity and Gemini

Key Takeaways

  • 1What an AI visibility audit covers
  • 2How to check your brand in ChatGPT, Perplexity and Gemini by hand
  • 3How to build a prompt panel for tracking
  • 4How to measure AI share of voice
  • 5AI referral traffic in GA4 and AI search data in Search Console

Quick Answer

An AI visibility audit is a structured check of how ChatGPT, Perplexity, Gemini and Google AI Overviews describe your business, whether they cite your website, and how often they recommend you instead of a competitor for the searches that matter. Run it with a fixed panel of 30 to 50 prompts, score every answer for mention, citation and accuracy, and repeat it monthly. Most Indian SMEs we audit score under 20% share of voice on their core category prompts, and that gap is what the audit exists to close.

This guide is for business owners and marketing managers in India who have noticed that enquiries now arrive with the words "I asked ChatGPT" in them. After reading it you will be able to run a first audit in an afternoon, know which numbers belong in a monthly report, and decide whether a wrong fact needs a content fix, a schema fix or a listings fix.

What an AI visibility audit covers

Classic rank tracking tells you where your page sits on Google for a keyword. It tells you nothing about what a buyer sees when they type the same question into an AI assistant and get a single written answer with three sources attached. An AI visibility audit measures four things across those answer engines.

  • Presence: does the answer mention your brand at all for prompts in your category and city?
  • Citation: when the engine lists sources, is one of them a page on your domain, and which page?
  • Accuracy: are the facts about you correct: services, location, pricing bands, founding year, people?
  • Share of voice: out of all the brands named across your prompt panel, what percentage of mentions are yours?

The audit is the measurement layer; generative engine optimization is the work you do afterwards. This post stays on the measurement side.

How to check your brand in ChatGPT, Perplexity and Gemini by hand

Before any spreadsheet, spend thirty minutes asking the engines about yourself from a fresh or temporary chat, using the same three prompt types: a direct brand question ("What does [brand] do?"), a category question ("Which companies make brass fittings in Jamnagar?") and a comparison question ("Who should I choose for payroll software in Rajkot?").

How to check your brand in ChatGPT

Open a temporary chat so memory and past conversations are switched off, and make sure web search is enabled, because without it the model answers from training data that may be a year or more old. Ask the direct brand question first and read the answer as a customer would: is the city right, are the services current? Then ask the category question and note every brand it lists and the order. Finally open the sources it shows and record which of your pages, if any, it used.

Perplexity brand check

Perplexity is the easiest engine to audit because it shows numbered citations beside each sentence. Run the same three prompts and, for every claim about your business, note the source number and the domain behind it. This is where you discover that the engine believes your turnover is "around 2 crore" because a five-year-old directory listing says so. A Perplexity brand check also tells you which third-party pages the engine trusts about your category, which becomes your outreach list later.

Gemini brand check

Gemini draws on Google Search and the Knowledge Graph, so its picture of you usually matches what your Google Business Profile and your website say. Ask the same prompts, then use the double-check option under the answer to see which statements Google can corroborate and which it cannot. Do a separate pass on Google Search itself with AI Overviews and AI Mode, because the app and the search surfaces often give different brand lists.

How to build a prompt panel for tracking

One afternoon of manual checks is a snapshot. To see movement you need the same prompts asked the same way every month, which is what a prompt panel is.

Prompt panel for tracking: 30 to 50 prompts in five buckets

  1. Brand prompts (5): "What is [brand]?", "Is [brand] reliable?", "[brand] pricing".
  2. Category plus location prompts (10 to 15): "ERP software company in Rajkot", "SEO agency in Gujarat for manufacturers", "best hospital management software India".
  3. Problem prompts (10): questions customers ask before they know they need you: "how to stop stock mismatch in a retail shop".
  4. Comparison prompts (5): "[brand] vs [competitor]", "alternatives to [brand]", "custom CRM vs packaged CRM for a small business".
  5. Buying prompts (5): "how much does a POS system cost in India", "which company to hire for a Flutter app in Gujarat".

Write the prompts the way customers actually phrase them and store them in a sheet with one row per prompt and one column group per engine.

How to track AI citations

For each prompt and engine, record three fields: whether your domain was cited (yes or no), the exact URL cited, and the position of that citation in the source list. After three months a pattern appears: two or three pages earn almost every citation, and they are rarely the ones you expected. Those pages deserve more internal links and quarterly updates.

How to monitor AI mentions every month

Answers from the same prompt vary between runs, so ask each prompt three times and record the majority result. Run the panel in the first week of every month, on the same day if possible, with the same settings (web search on, temporary chat, same country). A plain spreadsheet with 50 rows and 12 columns, filled by one person for two hours a month, is a working monitoring system.

How to measure AI share of voice

AI share of voice is the percentage of brand mentions across your whole prompt panel that belong to you. If your 40 category and buying prompts produce 160 brand mentions across three engines and 24 of them are yours, your share of voice is 15%. Track two more numbers alongside it: the first-mention rate (how often you are named first, which is where the click goes) and the citation rate (how often your domain is a source, regardless of whether you are named).

Competitor AI visibility: who is being named instead of you

The same spreadsheet gives you competitor AI visibility for free. Tally every other brand the engines named and sort by count. For the top three, look at why they appear: a strong review profile, an industry association listing, a mention in a widely cited article, or simply a clearer services page. You are finding which trust signals the engines weighted and which ones you lack. Keep competitor names in your internal sheet, not in anything you publish.

AI referral traffic in GA4 and AI search data in Search Console

Prompt panels measure what the engines say. Your analytics measure what buyers did after reading it.

How to see AI referral traffic in GA4

Open Reports, then Acquisition, then Traffic acquisition, and change the primary dimension to Session source. Filter for sources containing chatgpt.com, perplexity.ai, gemini.google.com and copilot.microsoft.com. For a permanent view, create a custom channel group called "AI referrals" that matches those domains. Expect small numbers, often 1% to 4% of sessions for an Indian B2B site, but check the landing pages: AI referrals land on deep pages and usually convert better than social traffic.

AI search data in Search Console: what is and is not there

Search Console does not give a separate filter for AI Overviews or AI Mode. Google has said impressions and clicks from those surfaces are counted inside the Web search type, mixed with ordinary blue-link results. So the signal you can read is indirect: look for queries where impressions are stable or rising while clicks and click-through rate fall, especially informational queries. That shape usually means an AI answer now sits above your result. Pair that list with your prompt panel to confirm whether the answer cites you.

How to score AI answer accuracy

Accuracy is the part of the audit owners care about most, because a wrong fact loses a lead silently. Score every brand answer on a four-point scale.

  • 3, correct: every fact matches your website and Google Business Profile.
  • 2, minor error: an old product name, a slightly wrong founding year, a missing service.
  • 1, major error: wrong city, wrong industry, pricing that is far off, your brand merged with another company of a similar name.
  • 0, fabricated: the engine invented an office, a founder or a client list.

Average the score per engine and per month. An average under 2.5 means your public footprint is contradicting itself somewhere and the fix below is urgent.

How to fix wrong information in AI answers

Start with the source, not the engine. Perplexity and Gemini will usually show you the page that carried the wrong fact; it is often an old directory profile, a press mention from 2019 or your own about page that nobody updated. Correct it at the source where you can, then publish one authoritative page on your own domain that states the facts plainly (what you do, where you are, since when, who runs it) and mark it up with Organization and LocalBusiness schema. Our schema markup guide for local businesses covers the exact fields. Answers that rely on live web search correct themselves within days to a few weeks; answers that come from training data only change at the next model update, so the goal is to make sure the live web version of you is unambiguous.

What we found auditing Rajkot and Gujarat businesses

Since early 2025 we have run this audit for SEO clients and for our own brand, and four patterns repeat.

A bearing manufacturer near Shapar scored zero mentions on 14 category prompts even though it ranked on page one of Google for most of them. The engines were naming companies that had detailed product pages with specifications in text, while its catalogue lived in PDFs. Rewriting six product pages as plain HTML with spec tables took the citation count from 0 to 9 within two months.

A dental clinic in Rajkot was being described by two engines with the wrong area of the city because its oldest listing, created by a previous marketing person, used a since-closed branch address. Correcting three listings and the about page fixed the Perplexity answer in eleven days.

A kitchenware exporter in the Rajkot GIDC belt had a 4% share of voice on export sourcing prompts. The brands winning were the ones with country-specific pages and a visible certifications section. Adding both lifted it to 19% over one quarter.

For our own brand, spelling variants of our name returned nothing in two engines until we added alternate names to our schema and about page. Entity work is cheap and it is always the first fix we recommend.

AI visibility report template

A monthly report should fit on one page and compare this month with last month. This is the layout we use.

SectionWhat it containsMetric to track
SummaryThree sentences on what changed and whyShare of voice, up or down
Presence by engineMentions across the panel for ChatGPT, Perplexity, Gemini, AI OverviewsMention rate per engine (%)
CitationsPages on your domain that were cited, with countsCitation rate, top 5 cited URLs
AccuracyAverage accuracy score and a list of wrong facts foundScore out of 3, open errors
Competitor visibilityTop 5 brands named instead of you (internal only)Their mention counts
TrafficGA4 AI referral sessions and leads attributed to themSessions, conversion rate
Search Console signalQueries with rising impressions and falling CTRCount of affected queries
ActionsWhat will be fixed before the next runOwner and date

Keep the raw prompt sheet as an appendix for the SEO team.

How to decide what to do after the audit

The audit will produce a long list. Fix accuracy errors first, because they cost leads today. Then fix the entity basics: consistent name, address, phone and alternate names across your site, your Google Business Profile and your listings. Then rewrite the two or three pages the engines already cite so they answer the category prompts completely, with specifications, pricing bands and a table. Only after that spend on new content or outreach. A typical Rajkot SME can clear the first two stages in a month for well under ₹50,000 if the website is in reasonable shape.

If you would like a fixed quote for a full AI visibility audit with a 50-prompt panel and a written fix list, send your website and your top five services through the contact page or on WhatsApp at +91 7265004040.

Tags

ai visibility auditcheck brand in chatgptperplexity brand checkgemini brand checkai share of voiceprompt panel for trackingtrack ai citationsai referral traffic ga4monitor ai mentionscompetitor ai visibilityai search data in search consoleai answer accuracyfix wrong information in ai answersai hallucination about my businessai knowledge panelai visibility report template

Need Help with SEO & Marketing?

Our team at JK Tech Hub is ready to help you build the right solution for your business. Let's discuss your project.

Contact Us