ChatGPT pulls results from Microsoft, from Google through scrapers and from Labrador, an index OpenAI built itself. Here is the evidence, checked on 2026-10-17, and what it means for your site.

We scored sixteen GEO agencies on eleven things a US buyer can check on their own pages, from prompt tracking and attribution to dated cases and price. Agenzy leads at 21 of 22, followed by Omniscient Digital at 20 and Seer Interactive at 19, with a source behind every cell.

Month one writes the tracked prompt set, measures the baseline per engine and opens the site to AI crawlers. Month two ships the pages and the off-site placements. Month three delivers a per-engine report against that baseline. This is our calendar, with the client input each row needs, the check that proves it landed, what day 90 is worth in Google demand and attributed leads, and a section on what 90 days does not buy.

One real AI visibility audit of one real B2B SaaS company. We have masked the company and every number is exactly as measured. Seventy-five buyer prompts, 2,153 AI answers across ChatGPT, Perplexity and Google AI Overviews, four markets, one day in July 2026. The company was named in 23.59% of answers and in 1 of 376 answers to problem-shaped questions.

Every GEO proposal offers an AI visibility audit, citation-ready content, technical work and digital PR, because that stack is identical across the market. We reviewed 22 agencies' public pages, directory profiles and press coverage in September 2026. This is what to ask instead, what a real answer sounds like, and the one request a relabeled SEO retainer cannot fill.

Nine of the ten things a GEO report should show, taken from one real monthly report for a real client, masked at the name level and exact at the number level: 15 prompts tracked daily across ChatGPT, Perplexity, Google AI Overviews and Claude, 4 May to 14 September 2026, with the eleven flat weeks at the start and the prompt still at 0% after four months. The tenth, leads by source, comes from a second company, and the one number the report refuses to print is a single composite visibility index.

In May 2026 Outcraft AI's robots.txt blocked four AI crawlers and the brand appeared in 5 of 100 tracked buyer prompts. This is the week by week method that followed: what shipped in each of fifteen weeks, how we checked that it landed, the five weeks when the line went sideways, and 98% share of voice on one tracked prompt in the 14 days to 2026-08-31.

llms.txt is a small text file that points AI tools to a website's most important pages. Server logs from 137,210 websites show that almost nothing asks for it, and Google says its search ignores it. What the file holds, who reads it, and the four fixes that matter more, with one client site where a Cloudflare setting kept OpenAI's bots out.

AI crawlers do one of three jobs: they collect pages for model training, build an AI search index, or fetch a page for a person's question. This reference lists 51 bots, who runs each one and what blocking it costs. It adds a month of bot traffic on one site, the robots.txt we recommend and the firewall check that robots.txt cannot do for you.

Eight numbers show whether AI assistants name your brand, use your pages and send you buyers. Each answers its own question, and a score combined across engines can more than double in a week while ChatGPT moves about one point. Read them per engine and against a baseline, and track only buyer questions that leave your brand name out.

LLM SEO, also called LLMO, is the work of getting your brand named and described correctly by the AI models behind ChatGPT, Gemini and Claude. A model answers from two places, what it learned in training and what it looks up live, and each needs different work. Paid work shows first in live answers, because a model's memory changes only with its next version.

Most of what AI search does for a business never arrives as a click, so last click analytics reports it as Google or Direct and the revenue looks like it came from somewhere else. This guide is the method we run on every client to measure it: traffic from AI models, visibility on the prompts buyers ask, and revenue confirmed by the buyer at two touch points. It includes the counting rules, a worked example and a checklist you can run on your own site in a week.

GEO stands for generative engine optimization: the work that gets your business named and cited when someone asks ChatGPT, Gemini or Google AI Overviews for a recommendation. Here are the four steps of the work, how an AI assistant picks the companies it names, what the research measured, and when we would tell a business to wait.

AEO, or answer engine optimization, is the work of writing pages that an AI can lift an answer from and name you as the source. Here is what an answer engine does with a question, which part of GEO the term covers, the two numbers it is measured by, and the cases where waiting is the better call.

AI SEO is the name many people use for getting your business named and cited in AI answers. This guide explains the work in four plain steps, what changes from classic search work and what carries over, and where a beginner starts. It also covers what the peer-reviewed research found, including the old habit that now measures worse than doing nothing.

The letters AIO are used for three different jobs, and two of them spend your money in different places. Here is who uses the word each way, since when, and the question to ask before you approve a budget.

Prompt tracking is mystery shopping inside ChatGPT, done every day by a robot. It asks your buyers' questions and records which brands the answer names, engine by engine. This is what it measures, why a single reading tells you nothing, and when a business should wait.

Every visibility number we report is measured over a prompt set, so the set decides what the number means. This guide is the method we use on every client to build it: four sources of buyer language, the shape a prompt has to take, the two tests every prompt passes before it is tracked, the client review, the freeze and the baseline that follows.

Six AI visibility tools priced on one unit, cost per prompt per engine per day, computed from their own pricing pages: Peec AI, Profound, Otterly.AI, Semrush, Ahrefs and Trendos, the Vilnius-built tracker with a free plan. Plus two that publish no price you can normalize, Scrunch and Passionfruit Labs, and who owns each tool after the Adobe and Sitecore deals.

ChatGPT has no ranking you can climb. It cites the pages whose passages answer a buyer's question, and most of what it reads sits on other people's sites. Four steps, in order, each with a check.

Agenzy analysed 2,283,923 AI citations from ChatGPT, Google AI Overviews and Perplexity across 138 business projects, most of them in Lithuania. Reddit and YouTube lead, no website holds more than 2.1% of citations, and the three engines share only 2 of their top 10 sources.

We are the most-named GEO agency in Lithuania and fourth in Estonia at 3.1%. Same brand, same method, three answers. Six European markets measured one way: our own Baltic numbers, three client markets in Greece, Poland and Spain, the sources ChatGPT names per market, the fan-out language data and the four checks to run on Germany, where we hold no client data.

On 2026-09-07 ChatGPT answered a UK shopper with three ramps, their prices and its reasons, and dropped a fourth because the maker's site said UK stock was out. The shops it names are the shops whose pages it fetched, with price and stock readable. Here is what that means for an e-commerce or DTC brand, with OpenAI's own criteria and a 90-day plan.

A GEO agency retainer costs 3,000 to 8,000 EUR a month for mid-market work, and premium specialists charge 7,000 to 20,000 EUR. We re-read every published price in this guide in October 2026 and show the bands by country, what each level buys, the costs outside the monthly fee and a way to compare two offers line by line.