Getting cited by AI search,
answered plainly.

Two things have to be true at once. The model has to be able to read your site, which means letting the search crawlers in and keeping your pages fast and plainly written. And it has to have a reason to name you, which comes from being described the same way across the places it already trusts: comparison posts, review sites, directories, forum threads, your own product pages. Backlinks count for less here than they do in Google. What counts is whether a model can find a consistent, quotable description of what you sell and who you sell it to. What is GEO goes through the mechanics.

Almost always because the model has more to go on about them. It has read pages that name your competitor inside your category, in the words a buyer would use, and it has read less of that about you. This is rarely a verdict on your website being worse. It means that when the question is asked in general terms, your name is not attached to the category strongly enough to come up. Getting described in the same places, in the same language, is the work.

Partly, and the two halves are worth keeping separate. OpenAI sells ad placement inside ChatGPT, which is bought and labelled like advertising anywhere else. The recommendation inside the answer itself is not for sale. Services that claim to sell it usually mass-post your brand onto low-quality sites, and that tends to backfire: models weigh where you are described and how consistently, so a burst of near-identical posts on sites nobody cites is a pattern they discount.

Only if you do not want your content used for training, and it is a separate decision from AI search. OpenAI runs two crawlers: GPTBot collects training data, and OAI-SearchBot builds the index ChatGPT reads when it answers a live question. Blocking OAI-SearchBot takes you out of ChatGPT answers entirely. Most brands that want citations block the training crawler and let the search crawler through. The crawler checker shows which ones your robots.txt allows today.

Not much. No major AI company has committed to reading the file, and Google states in its own documentation that it ignores it. Publishing one costs nothing and does no harm, but it is not a control mechanism and it will not get you cited. robots.txt is the file the crawlers actually honour, so that is the one worth getting right.

Both, and start with what you have. Pages that already rank often lack only the thing a model needs in order to quote them: a direct answer near the top, a specific number, a clear line about who the product is for. That is cheap to add and it moves first. New content is for the gaps, which in practice are comparison pages and the questions your category asks that you have never written about.

Ask the engines the questions your buyers ask, and write down what comes back. Not your brand name, which will always find you, but the general question: the best tool for a job, who to hire in a city, which option suits a situation. Run it in every engine your buyers use, and run it repeatedly. Sourso does this as a free AI visibility check and keeps the result in your workspace.

Because the models are not deterministic, and because many questions send them to the live web. The same prompt can pull different sources an hour apart, and rewording the question slightly changes the answer more than most people expect. This is why one screenshot proves very little, and why visibility is reported as a rate across repeated runs rather than as a yes or no.

Thirty to fifty for one brand in one market is the working range. Below that, a single answer swings the number and you cannot tell a real change from noise. Far above it, you are mostly paying to ask the same thing in different words. Spread them across the stages a buyer moves through: the general category question, the comparison question, and the question that names you directly.

Sourso is a personalized AI CMO for GEO, backed by an agent marketing team that tracks how AI answer engines describe your category, writes the content that gets you cited, and reports what changed. You brief it in chat and you approve what ships.

Ten lines of work so far, all listed on professions. It began with researchers, founders and small marketing teams, and the same problem turned out to look identical for dentists, real estate agents, insurance brokers, lawyers, consultants, financial advisors and cross-border sellers. In every one of them, buyers now ask an AI who to go to before they contact anyone, and whoever gets named gets the call.

No. Describe your brand and your competitors once, and the Source, Solve and Operate teams handle the practice from there. What you do is read what they found and approve what they wrote.

Everything is free during the beta. After that your AI CMO is 29 US dollars a month or 290 a year, and specialist marketplace agents are hired separately by the job, against a budget you control. The AI visibility check stays free and does not ask for a card. Full pricing.

A specialist outside your built-in team that you can hire in the middle of a conversation, paid by the job. You see the scope and the price before you approve it. Every hire runs sandboxed, is scoped to one identity, and does nothing without your approval. Browse the marketplace.

Yes. Each identity in a workspace keeps its own signals, drafts and dashboards, and every marketplace hire is scoped to a single identity. Workspaces and identities has the detail.

Not yet. The marketplace runs our own agents for now. Build an agent describes the flow we are building and the manifest fields it will ask for, and it will be announced in the changelog when it opens.