Straight answers about our data

Where the answers come from

We measure answers as real users see them. For ChatGPT, Gemini, Google AI Overviews, Google AI Mode and Perplexity, AEOHIT works with industry data providers that capture what those consumer interfaces actually show people. For Claude and Grok, we ask through Anthropic's and xAI's official APIs with web search turned on. Those answers come from the API, not the consumer Claude or Grok apps, and can differ from what the apps show.

Perplexity is captured as users see it, through an industry data provider’s consumer browsers. When that capture is unavailable, we ask Perplexity’s official API with web search turned on instead. Those answers are marked “via official API” in your dashboard and can differ from what the Perplexity app shows.

Microsoft Copilot: Capture temporarily unavailable — not counted in your score and not charged. Its data provider can no longer capture its answers, and there is no official Copilot API we could ask instead. We will measure it again as soon as a capture works, and say so here.

In your first 3 days of Pro, the baseline measures your top 25 questions on six engines — every engine except Grok, which joins from your first paid scan. Every paid full scan measures all seven.

We have no access to any engine’s private systems, and we do not pretend to. Every answer in your dashboard shows which engine gave it and when.

Markets we measure

We measure answers as customers in the United States, the United Kingdom, the United Arab Emirates, Canada, Australia, Singapore and India see them. You choose the market for each brand and for the free check, and every question is asked from it: the data providers capture each answer from that country, and the official APIs are told the person asking is there. Outside India, questions are asked in English.

City-level targeting, a question asked as someone in Pune or Chennai would ask it, exists for India only, across 40 cities. In every other market we ask from the country as a whole. We do not measure other countries yet, and we never present one market’s answers as another’s: the free check’s report and your dashboard name the market their answers come from.

How often we ask

Every question you track is sampled at least once a week; higher plans sample more often. Each time, we record the full answer, every brand it names, where you appear, and which sources it cites.

What each plan measures

Every plan measures all seven engines. ChatGPT, Gemini, Google AI Overviews and Google AI Mode answer every question you track, on every plan. Perplexity, Claude and Grok answer your highest-intent questions, the ones closest to a purchase (“best dentist in Pune” ranks before “how to choose a dentist”): on Pro, Perplexity and Claude your 25 highest-intent questions and Grok your 10; the table shows every plan.

Common questions in your market, the ones every business in your category is asked about, are answered once a week and that answer is shared, so every business in a market is compared on the same answer. Questions about your own brand are asked for you alone. In a heavy month the engines that answer part of your questions may focus on your highest-intent ones for the rest of the month; your dashboard says so whenever it happens ("Measured on 13 of your 50 questions this week"), and your baseline scan, the weekly scan on the other engines and the fixes already written are never reduced.

This table is the current service specification. We may adjust engines, measurement frequency and allowances per plan to keep the service reliable; the table here always shows what is in force.

Why answers differ

AI engines can answer the same question differently for different people — by location, language, account history and the day you ask. We do not reproduce any one person’s personalised results. That is why we sample on a schedule and score the trend, not a single answer.

Indian languages

Questions in Hindi, Hinglish, Tamil, Telugu, Marathi and Bengali are written the way people ask them in that language, not translated from English, and each one goes only to engines with evidence that they answer in that language: a vendor or data-provider language list, a launch announcement, or a general-purpose assistant that replies in the language it is asked. English, Hindi and Hinglish go to all seven. Tamil, Telugu and Marathi go to ChatGPT, Gemini and Google AI Mode, and Bengali also to Claude: Google has launched AI Overviews in India in English and Hindi only, and we have found no published support for them from Perplexity or Grok, or from Claude beyond Hindi and Bengali. A pair we do not run is shown as not measured, never as a miss, and stays out of your score. We add an engine to a language once a live check shows it answers.

The score

Your AI Visibility Score combines presence (how often you appear), prominence (how early), share of voice (you against competitors), sentiment, and citations (whether engines cite your site as a source). Each engine is weighted by its real-world referral share, and the weights are recalibrated as the market shifts.

The site audit

The site audit reads the pages your sitemap lists (without one, the pages your home page links to), through a polite crawler that obeys your robots.txt, and checks what AI engines need to read, trust and quote you: an llms.txt file, crawler rules that let AI engines in, a sitemap, clear titles and descriptions, one main heading, valid structured data, a direct answer near the top of each page, FAQ answers and a consistent business name. Its SEO essentials check (canonical tags, noindex mistakes, broken pages, duplicate titles, mobile viewport, image alt text, plain-http links, Open Graph tags, render-blocking scripts) has a score of its own and never changes your AI visibility score.

Apps

If your business has a mobile app, add its App Store or Google Play link. AI engines learn about an app from web pages: its public store listing, your own site, and what reviewers and the press write. That is what we check. We read each listing’s public page once per site audit, from the store of your market, through the same polite crawler that reads your site (it follows each store’s robots.txt), and check seven things: that your site links the listing, that the developer name shows it is yours, that the first 300 characters of the description say what the app does and for whom, how many ratings it has, that your llms.txt lists it, the App Store banner tag and your Android app links. A listing is flagged as thin on ratings below our own threshold, scaled to the size of each market and not a store rule: fewer than 1,000 ratings in the United States, 300 in the United Kingdom, 100 in the United Arab Emirates, 200 in Canada, 200 in Australia, 100 in Singapore and 1,000 in India. These checks have their own App listings score and never change your AI visibility score.

In answers, “Named as an app” counts the answers that cite your store listing as a source; it is shown beside your score, never inside it, and re-measured with your fixes. We never install or test the app, never ask for App Store Connect or Play Console access and offer no keyword (ASO) tools. A rating we write into a fix is copied from your listing with the date we read it, never estimated, and no store ranking is promised.

Local visibility

For a business people visit, add its address in Settings (or let your AI business page place it). Each week we ask Google Maps the local questions people nearby type — your own tracked questions that name your town or say “near me”, then questions like “dentist in Pune” or “best dentist near Koramangala” and your own services in your area — and record where your business appears on the Maps list and which businesses are shown first. We read your Google listing (its categories, hours, phone, website, description, photos, rating, latest reviews, posts and questions and answers) and check the directories that matter in your market and category — Bing Places, Apple Maps where search can find it, Yelp, Tripadvisor, Justdial and Sulekha in India, and others — for your listing and whether its name, address and phone match yours. The data comes from industry data providers, on the weekly cadence; once you connect your Google Business Profile, your listing is read from Google itself.

The Local visibility score weighs your presence on Google Maps for those questions (half), how complete your listing is (30 %) and your reviews — rating, how recent and how many, against the businesses shown next to you (20 %). A part we could not read that week is left out and the score says so; it sits beside your AI visibility score and never changes it. A directory listing that search results do not show is reported as “not found in search results”, not as missing, because a page search has not indexed can still exist.

We cannot post reviews, change ratings or edit your Google listing for you. We write the description, services, answers, posts, review replies and directory details from your own words for you to paste; publishing to Google directly needs Google’s approval of our app and your connection, and every change waits for your approval. Nobody controls Google Maps rankings, and we promise none.

No ranking guarantees

Nobody controls AI engines. We measure honestly, write the fixes most likely to get you cited, and show you the trend. We will never promise a placement — anyone who does is guessing.

What we do with the data

Your tracked questions and the answers they produce are used to score your visibility, find where you are missing, and write your fixes. Answers to the same question are shared across customers in an anonymised measurement pool, so each question is asked once and prices stay low. Your brand settings, scores and account details are never shared, and we never sell data. Details are in our Privacy Policy.

How long we keep it

Raw answer text in the shared pool is kept for 30 days. Your scores and trends stay for as long as your account is open. When an account is deleted, its personal data is purged within 72 hours, except invoices we must keep by law.

Where the prices we compare come from

When we compare our prices with agency retainers or other tools, we link to the published source and the date we checked it. If a source cannot be linked, we do not use the comparison.

• First Page Sage, “GEO & AEO pricing statistics”, updated 7 Oct 2026
• upGrowth, “GEO and AEO pricing benchmark, India 2026”, 20 Apr 2026
• Storyboard18, “Brands pay Rs 1 lakh to Rs 15 lakh monthly for AI visibility in India”, 1 Oct 2026
• Competitor plans and prices: each comparison page links to the vendor’s own pricing page and states the date we checked it.

Sample data is always labelled

When AEOHIT runs on sample data — for demos and testing — every page carries a SAMPLE DATA banner and every sample answer is marked.