The questions people ask first

SEO and AI visibility software for businesses and agencies in India and worldwide. It runs a twelve-week campaign for one business: the searches people in its town make, the fixes its site needs, new and better pages written from its own facts, and a weekly check of Google, Bing and the AI answers.

Both. Most of the work is ordinary SEO: searches, rankings, the site and its pages. The AI assistants answer from the same pages and listings, so Oftheard checks what they say every week as well.

A business, in India or any other country, that wants more customers to find it, and agencies that do that work for clients.

The work of getting named by an AI assistant when someone asks it what to buy or whom to call. It is spelled optimization in most of the world, and also called generative engine optimisation (GEO); the two names mean one thing. Search engines rank pages; answer engines such as ChatGPT, Gemini, Perplexity and Google’s AI Overviews name a few businesses and leave the rest out. Being one of the names rests on two things: the assistant being able to read your site, and your site, and the pages it reads about you, plainly stating what buyers ask about. AEO vs GEO explains the terms.

SEO earns a place on a list of results; answer engine optimisation earns a mention in a single answer. On Google, a page at position six still gets clicks. In an AI answer, a business that is not named does not exist for that buyer. The two overlap more than they differ: the assistants that search before answering read the pages that rank, and they read directories, reviews and comparison pages too, so where you are named matters as well as what you publish. That is why an Oftheard campaign does the SEO first and checks the AI answers alongside.

Yes. Both mean how often a large language model, such as ChatGPT, Gemini or Claude, names or recommends your business when a buyer asks it something in your field. Some tools call it AI share of voice. Oftheard counts it plainly, per assistant and per question: of the AI answers to the questions your campaign tracks this week, how many named you, which pages each answer used, and who was named instead.

Usually for one of four reasons, and each needs different work. The pages the answer used do not mention you, so the work is getting named on them. The assistant does not find your page when it searches, so the page has to rank. Bigger names hold the answer, which takes other sites linking to you, and time. Or your page does not say it the way the answers do, so it needs your buyers’ words. The weekly report gives the main reason for every question where you were left out.

Because the search behind the answer returns different pages each time, and the model chooses its words afresh. Ask the same question twice and the names can change, so one answer is not a measurement. Oftheard asks the same questions every week, on the same assistants, from the same places, and keeps every answer whole with the pages it used. Read the trend across weeks and the count across many questions, not one answer against the last. There is a longer note: why AI answers differ.

Weeks for the assistants that search before they answer, longer for anything that rests on what a model learned in training. Perplexity, ChatGPT with search and Google’s AI Overviews read pages when they answer, so a new page or a new listing can change an answer once it is indexed and ranks. Asking an assistant again teaches it nothing; publishing the page it needs to read does. Searches about your town or a narrow service often move in one to three months; national ones take other sites linking to you, and time.

It is cheap and cannot hurt, but it is not what gets you recommended. An llms.txt file is a plain-text summary of a site for AI crawlers. Some assistants read it; none is known to rank on it. The site check looks for it because a missing one is a missed chance to say what you are in your own words, and oftheard.in publishes its own. More in does llms.txt work?

The major ones do, and many sites block them without meaning to. A rule written years ago to keep unknown bots out often turns away GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot or Google-Extended today, and some hosts and firewalls turn them away before robots.txt is even read. The site check reads your robots.txt, fetches your pages as the AI crawlers do, and says which are refused, because a crawler that is refused cannot read the page that would have got you named.

No. The score is a ceiling, not a prediction. A site an assistant cannot read cannot be recommended, so a low score holds you back; a high score only means nothing on your side is stopping it. Whether it then names you is what the weekly report measures. oftheard.in’s own site scores 100, and in week 6 of its campaign it was named in 3 of 45 AI answers.

No one can honestly promise that. Searches about your town or a narrow service often move in one to three months. City-wide searches also need reviews and listings; national searches need other sites linking to you, and time.

One hundred checks in five groups. Being found by AI: whether the AI crawlers are let in and get the page, llms.txt, text that reads without JavaScript, answers a sentence can be lifted from. Being found on Google: robots.txt, the sitemap, titles, headings, structured data, speed and phones. Your location: whether the address, hours and Google listing agree, scored only for a business with a place. Then AI assistants doing things, and general housekeeping. It takes about half a minute and changes nothing on your site. Every check, explained.

ChatGPT, Gemini, Perplexity, Google AI Overviews and Google AI Mode every week, and Claude every other week on Growth. Your questions are asked as a buyer in your own country, and each answer is kept whole with the sites it read.

No. Copilot is not one of the assistants asked. What a campaign does measure is Bing, whose index Copilot answers from: your Bing position for every tracked search, every week. The site check has a Bing index item that says whether Bing has your pages at all; if it does not, submit the site in Bing Webmaster Tools. The crawler sensor also counts Bingbot’s visits, and the people who arrive from Copilot.

Not itself. Nothing here signs in to ChatGPT or drives a browser of ours. The ChatGPT and Gemini apps, and Google’s AI Overviews and AI Mode, are read as a buyer sees them through a data provider, DataForSEO, named in the privacy policy. Perplexity is asked through its own API with search on, and Claude with its web search on. Every answer is kept with the assistant it came from, so assistants are never mixed into one number.

Because the answers differ by country. Asked from India, an assistant cites Indian directories, rupee prices and Indian review sites, and names different businesses than it would for a buyer in the United States. So each question is put to the assistants as a buyer in the business’s country, which for a business in India is India, and Google is searched from each question’s own place: a city, a state or the whole country. The answer you see is the one your buyers get.

Yes, in a campaign. You name the businesses you compete with, up to 5 on Starter and 10 on Growth, and each campaign measures two to five of them on the same questions as you, every week: whom the AI answers named instead of you and for what, and which sites hold Google’s top ten. Names the answers recommended that are on nobody’s list are shown too, so a rival you did not know about still turns up. How competitors are measured.

Yes, with a campaign. The site check says whether the door is open; the crawler sensor says who came in. Install a WordPress plugin, a Cloudflare worker or a few lines on your own server, and it reports every visit by GPTBot, ClaudeBot, PerplexityBot, Googlebot and the rest, page by page, kept for ninety days. Each page the campaign publishes shows its reads before and after it went live. Nothing about your human visitors is sent. The install guide.

From your own site, read for you, then confirmed by you: prices, services, places, contact details and proof. A draft that says anything else is flagged before you see it.

Google tests a changed page for about three weeks, moving it up and down. The campaign waits that out, and opens an update only once a page has really slipped.

Yes. Each question is searched on Google from the place you give it, a city by its point and a few kilometres round it, so you see the results a buyer in your town sees. Each week shows on how many questions Google showed a Maps box, whether you were in it, and who was. The site check covers the local signals: the address and hours in the page markup, whether your Google listing agrees with the site, and whether the pages name the town. The AI assistants are asked from your country, not your town.

Yes. At setup you choose the language the questions are asked in: English, or Hindi, Tamil, Telugu, Kannada, Malayalam, Marathi, Bengali and others. The language goes with every Google and Bing search and every question to the assistants. Questions are not translated for you, so type them the way your buyers would; and a business name written in another script may not be recognised in an answer.

Oftheard is built in India and works for a business in any country; its prices are in rupees. Each business has a country, set at setup: the AI assistants and Bing are asked from it, and Google from up to ten places a campaign picks, each a city, a state or a whole country. The questions can be asked in English or in another language of that country. Places and languages.

No. Every draft comes as HTML to paste into any site. On WordPress, an approved draft is published in one press and undone in one press.

They are left alone, because replacing them could break the design. The draft is saved in WordPress as a separate draft page for you to place.

You do, or the business owner when an agency runs the campaign. Nothing reaches your site until a person approves it, and the old page is kept so it can be put back.

Oftheard uses a WordPress Application Password, which you can revoke at any time. It is stored encrypted and never shown again, and only the approved page is ever changed.

You do. Everything a campaign makes, answers, rankings, drafts and reports, stays with the business on your account; for an agency’s client, the agency and the owner both see it. It is kept in Google Cloud in Mumbai (asia-south1), never sold, and never used to train any model. Delete a business, or the whole account, from settings and it is deleted at once, or write in and it is done within thirty days. The privacy policy names every service that handles it.

There is no connector to them. What a campaign connects to is Google Search Console and Google Analytics, read-only, for your clicks, impressions and visits and the searches your site already shows up for; WordPress, to publish approved pages; and a read-only MCP server, so Claude Code, Cursor, VS Code or Claude Desktop can read the campaign and the site fixes. The MCP server reference.

The site check, with no account and no card: all hundred checks in about half a minute, each problem with its fix. A free account adds one free report on up to three questions, once, not every month: those questions checked on Google and in the AI answers. An agency account trying Oftheard gets three, one each for three clients. Weekly runs, competitors, page drafts, publishing, the crawler sensor and the MCP server come with a campaign.

The site check is free. A twelve-week campaign is ₹4,999 on Starter or ₹9,999 on Growth, per business, paid once. Full pricing.

Yes. One account holds all your clients, each on its own plan, and each owner agrees to the campaign by a code sent to their business email.

You get the before and after, search by search, and what to do next. Extend for another twelve weeks, or stop.