GEO stands for Generative Engine Optimization. It is the work of making your business easy for AI engines to find, understand, and cite. ChatGPT, Claude, Perplexity, Microsoft Copilot, and Google's AI answers all write their responses from sources they trust. GEO is how you become one of those sources. This guide explains it in plain English, using real numbers from a real shop.
The engine details below were checked in July 2026. Crawler names and reporting tools in this field change month to month, so verify against each provider's documentation before acting.
Search behavior is splitting. People still Google things, but a growing share of questions now go straight to an AI: "how do I fix my product photos," "what should I look for in a memory quilt," "is print on demand worth it." The AI answers in one block of text, names a few sources, and most people never see a list of links. If your business is not readable to those engines, you are not in the room where the answer gets written.
SEO earns a ranking on a results page, and the customer clicks through to you. GEO earns a citation inside a written answer, and the customer may never click at all. That changes what you optimize. Rankings reward authority and links. Citations reward clear, self-contained passages that an engine can lift and attribute. The disciplines overlap heavily, and the good news is that the same clean foundation serves both.
On my mom's Shopify boutique, two clean guide pages earned more than 80% of the shop's AI citations in a month. Not the homepage. Not the product pages. Guides that answer one real question end to end, written in short passages with descriptive headings. AI answers are assembled from passages. Pages that read like clean passages get pulled; pages that read like brochures get skipped.
Most GEO advice treats "AI engines" as one thing. They are not, and the differences change what you can do and what you can prove.
Microsoft Copilot is the one I can measure on my mom's shop, which is why almost every number I publish comes from it. Copilot leans on Bing's index, and both Bing Webmaster Tools and Microsoft Clarity report citation counts and the grounding queries behind them, free. If you want a scoreboard that names the queries, this is where it currently comes from.
Perplexity shows its sources inline on nearly every answer, which makes it the easiest engine to audit by hand. Its crawler is PerplexityBot, and it also runs Perplexity-User, which Perplexity's documentation says generally ignores robots.txt because a person asked for that specific fetch. So robots.txt is a partial control here, not a complete one.
ChatGPT browses selectively rather than on every answer. OpenAI splits its agents by job: GPTBot for training, OAI-SearchBot for appearing in ChatGPT's search results, and ChatGPT-User for the live fetch made while answering. Those settings are independent, which matters enormously and almost nobody says it. There is no owner-facing citation report.
Claude splits the same way, with ClaudeBot, Claude-User and Claude-SearchBot. No citation dashboard for site owners. It is the engine I see discussed least in GEO content, which probably makes it the least contested.
Google is two things and the difference matters. AI Overviews sit inside Search, so they draw on pages that already rank and are governed by the ordinary snippet controls: nosnippet, max-snippet, data-nosnippet. Google-Extended is a separate control for Gemini apps and Vertex AI grounding, and Google states it does not affect Search inclusion. As of mid-2026 Google has also begun rolling out generative AI performance reporting inside Search Console, showing impressions from AI surfaces. It is a staged rollout and plenty of properties do not have it yet, including my mom's shop, so check whether yours does before assuming you can measure this.
The scoreboards are not equal, and the difference decides where to put your effort.
Bing Webmaster Tools and Microsoft Clarity report citations: how many times your pages were used, and which queries the engine was grounding on. Search Console's newer AI reporting shows impressions from Google's AI surfaces, which tells you that you appeared without telling you what you were quoted for, and not every property has it yet. ChatGPT, Claude and Perplexity report nothing to site owners at all.
So the defensible position is narrower than the one you usually hear: one engine family currently hands you citations and queries, one is beginning to hand out impressions, and three hand you nothing. Anything you are shown for those three came from sampling, which is a legitimate method when the run count is disclosed and a marketing number when it is not.
Fix your entity, meaning your name, address, and phone are byte-identical everywhere they appear. Write pages that answer one question each, cleanly. Add structured data that matches your visible text exactly. Allow the right AI agents in robots.txt, checking the search and fetch agents rather than only the training crawler. Fix your catalog data, especially alt text and file names. And measure from day one with free tools: Search Console, Bing Webmaster Tools and Microsoft Clarity.
Two things the sales pages never tell you. First, citations do not automatically mean traffic; the shop's citations sent almost no clicks. The win today is position for where search is going, and visitors tomorrow morning are not part of the deal. Second, being cited on how-to questions is not the same as being recommended on "where should I buy" questions. That commercial query is a harder, separate fight.
Pick the three questions your customers actually ask you in person. Answer each one on its own page, plainly, the way you would say it across a counter. Then check your robots.txt is not blocking the AI agents you actually want, and install Search Console and Clarity so you can see what happens next. That is a real GEO foundation, and it costs nothing but an afternoon.
No, though they share most of the same foundation. SEO optimizes for a position in a list of links, and GEO optimizes for being quoted inside a written answer. The overlap is large enough that good SEO work is rarely wasted on GEO. The difference shows up in how you structure a page: rankings tend to reward depth and authority signals, while citations tend to reward short self-contained passages an engine can lift without losing the meaning.
No, and this is the most common robots.txt mistake in this field. GPTBot is OpenAI's training crawler. Appearing in ChatGPT's search results is governed by OAI-SearchBot, and the live fetch made during an answer comes from ChatGPT-User. OpenAI's documentation states these settings are independent and gives allowing OAI-SearchBot while disallowing GPTBot as its example. If you want to keep your content out of training but stay visible, that is a supported configuration.
No. Google-Extended governs Google's Gemini apps and its Vertex AI grounding, and Google's documentation states it does not affect inclusion in Google Search. Since AI Overviews are part of Search, blocking Google-Extended is not the lever there. It is still a real decision, because it does remove you from Gemini and Vertex grounding. Make it deliberately rather than by leaving a plugin default in place.
It depends what you mean by measure. For citation counts with the grounding queries attached, Microsoft Copilot and its partner surfaces, through Bing Webmaster Tools and Microsoft Clarity. For impressions from Google's AI surfaces, Search Console's newer generative reporting, which is rolling out gradually and which many properties including my mom's shop do not have yet. For ChatGPT, Claude and Perplexity there is no owner-facing report at all, so any number came from sampling. Ask for the run count.
Allow PerplexityBot, then use the advantage Perplexity gives you: it shows its sources on nearly every answer. Ask it the ten questions your customers actually ask, in a fresh incognito window, and write down which sites it cites. You now have a named list of exactly who you are competing with, straight from the engine. Perplexity-User, the agent behind a user-requested fetch, generally ignores robots.txt, so your control there is partial.
Mostly no, and you should plan for that. On my mom's shop the AI citations drove well under 1% of sessions. Being cited and being visited are two different scoreboards and I track them separately. The case for doing this work now is position, ahead of traffic. If someone promises you visitors from AI citations, ask them for their referral numbers.
I cannot give you a clean answer from one store, and you should be suspicious of anyone who gives you a confident one. What I watched happen: the shop had a handful of scattered citations over seven weeks before the work, then the rate rose substantially in the six weeks after the catalog and guide pages were fixed. That is one shop, one niche, a small sample. Measure your own baseline before you change anything.
For the foundation, no. Everything in the six moves above is work an owner can do in a few afternoons, and I publish the steps free because the plumbing should not be a paid secret. I do take on a small number of audits, so treat that as my bias when you read anything here. The part that genuinely takes outside help is usually getting named in sources the engines already trust, because that one is slow and relational rather than technical.
In practice they overlap almost completely. GEO describes making your business visible inside the answers AI engines generate. AEO describes structuring your content so an answer engine can quote it directly. Both come down to letting the right crawlers read you, saying plainly who you are and what you sell, and being named where the engines already look.
On the evidence I have, no. My mom's shop has a Domain Rating of 4.9 and six dofollow referring domains, two of them flagged as spam directories, and Ahrefs shows exactly one page on her whole site with any external links pointing at it. That page is the homepage, and the homepage is not what gets cited. The two guide pages earning most of her AI citations have no backlinks at all. That is one store, so treat it as an existence proof: it shows authority is not required, and it does not show that authority never helps. Checked 28 July 2026.
I document this whole method on a real store, receipts included, on YouTube and in the articles. The deeper dive on checking your own store is here: Can AI read your store? The 10-minute check. If you would rather take the whole site in the order to use it, start here.