Generative Engine
Optimization Agency

Generative engine optimization (GEO) is the practice of getting a brand cited inside the answers that generative engines write — ChatGPT, Perplexity, Google AI Overviews, Gemini and Claude. Classic SEO competes for a position in a list of ten blue links. GEO competes to be one of the handful of sources a model reads, trusts and names while it composes a single answer. The unit of success is a citation, not a rank.

No, and any agency that sells it that way is selling you a rebrand. The two share the same foundation: a site an engine can crawl, an entity it can identify without ambiguity, and content that actually answers a question. What changes is what you optimise for on top of that foundation — self-contained passages, structured data, and third-party mentions rather than link volume alone. We run them as one service because splitting them would mean auditing the same site twice.

With a documented baseline and a repeated set of prompts. Before touching anything we record where the brand stands today: which queries it appears in, which it does not, and which competitors the engines name instead. Every month we re-run the same prompts and the same searches and report the delta against that baseline. Every number in the report carries its source and its date. A measurement we could not take is reported as not measured — never as a zero, and never as an estimate.

It depends on how well the engines already recognise the brand as an entity, and we will tell you which case you are in after the audit — not before. A brand with an established name, an active profile footprint and existing coverage moves faster than one whose name is shared with six other companies. What we commit to is the cadence of the work and the honesty of the report, not a date on the calendar.

ChatGPT, Perplexity, Google AI Overviews and Gemini, plus classic Google results for the same queries. We track them side by side because they disagree more often than people expect: a brand can be the first recommendation in one engine and absent from another for the same question. Which engines matter most is a decision we make with you, based on where your buyers actually search.

Yes. August operates out of Miami, Buenos Aires and Madrid, and two of those three markets are Spanish-speaking. Generative engines answer in the language of the question and draw on sources in that language, so a brand that only publishes in English is invisible to a Spanish prompt. We audit and build in both.

August does not publish prices, and anyone quoting one for our services is not quoting us. What we can tell you is what moves the number: how many questions you want tracked and how often, whether the work is monitoring only or monitoring plus content, and whether it includes the part that happens off your own site. Third-party pricing guides published in 2026 put agency retainers anywhere from about US$1,500 to more than US$50,000 a month — a range that wide is a sign the label covers very different work, so ask which of the three you are buying before you compare two quotes.

No, and neither can anyone else. A generative engine composes each answer at request time from passages it retrieves, and nobody outside the model controls that selection. What can be committed to is the work and the measurement: a documented baseline before anything is touched, a fixed battery of questions run on a fixed cadence, and a report that shows the readings that did not move alongside the ones that did. If a supplier guarantees a citation, ask them to put the mechanism in writing.

Ranking well and being cited are different outcomes produced by different systems, and one does not carry over to the other. A page can hold position three for a query and never appear in the assembled answer above it, because the engine retrieves passages rather than ordering documents. The way to find out is not to argue about it: pick ten questions your buyers actually ask, ask an assistant, and read who gets named. If it is you, this is not urgent.

Read access to Search Console and analytics, the ability to publish changes to the site or someone who can, and one person who can approve content. That is the minimum. What helps most beyond that is a list of the questions your sales team actually gets asked, because the tracked battery should be built from real buying questions and not from keywords that look good in a deck.

You do, in both cases. The accounts are opened in your name or you are made owner of the ones we open, the content lives on your domain, and the measurement log is yours to keep. This is worth asking every supplier before signing, because an engagement where the tracking lives in the agency's account produces a baseline you cannot take with you.

On the evidence available, very little. An Ahrefs study of 137,000 sites found that 97% of published llms.txt files received no requests at all during May 2026, and that most of the bots that did fetch one were SEO audit tools rather than AI engines. We publish one because it costs nothing and describes the company accurately, but we do not sell it as a tactic and you should be sceptical of anyone who does. What the file cannot fix is the thing that usually blocks a brand: nobody outside its own domain mentions it.

In practice yes, and the industry uses the two labels interchangeably — several agencies say so on their own service pages. GEO is the term the original research used; AEO is the one that spread through marketing teams. Treat a supplier who insists they are different disciplines with the same caution as one who has invented a proprietary acronym.

Then that is the first thing to fix, and no amount of content will substitute for it. Google reports this state as "Discovered — currently not indexed": it knows the URL exists and has decided not to fetch it, which usually means the domain has not earned enough crawling to be worth the request. It is diagnosable in minutes in Search Console, and the fix is rarely on the page itself. We know this one from our own domain.

Monthly, against the baseline documented at the start, and with the whole answer recorded rather than a yes or no: which sources the engine cited, what kind of page each one was, and whether it named companies at all or only directories. Those three move before a citation does, which is what makes them worth reporting. Readings that did not move are published too — a measurement that only appears when it is flattering is not a measurement, it is a selection.