Myths about GEO, that is about optimisation for generative search engines, have arisen in large numbers in a short time, and most of them do not hold up under scrutiny. Distinguishing between them saves both money and time. The first myth is that classic SEO has stopped working. The opposite is closer to the truth: generative systems draw on search indexes, so a page that is not accessible, indexed and technically sound stands no chance even in the answers. What has changed is what is being optimised – from rankings to passages – not whether the technical and content side needs to be in order. The second myth is that you can pay for placement in the answer. Advertising formats within AI answers are appearing and will keep increasing, but they are labelled as advertising and operate alongside the organic part of the answer, not instead of it. Offers that promise paid inclusion among a model's recommendations are either a misunderstanding or a scam. The third myth is the existence of a universal technical solution. Files and markup intended to tell models what to think of you are useful at most as a supplement. None of them replaces content that answers questions, nor a consistent identity across sources. A company that deploys a file and changes nothing else will not see any difference. The fourth myth is speed. A change in content does not show up in the answers immediately. Some systems draw on an index that refreshes over days to weeks, some on stored answers, and some on what the model learned during training, which may not change at all until a new version arrives. When correcting an inaccuracy about a company, you need to allow for a transitional period during which both versions coexist in circulation. The fifth myth is that it is enough to write longer texts. Systems select short passages, not entire articles, and a long text without specific data reduces the likelihood that anything can be extracted from it. What matters is information density, not length. The sixth myth is that generative search will kill website traffic. Some informational queries genuinely end without a click-through, and this is already happening. At the same time, however, the value of the visits that do arrive is rising – they come later in the decision process, after comparison, with the specific intent of verifying a piece of information. The right response is not to deny access to robots, but to prepare pages for a visitor who is already eighty per cent decided. The seventh myth is measurability using a single precise metric. Answers differ on every run even without any intervention on your part, so a single measurement proves nothing. What makes sense is a fixed set of questions, regular measurement and trend evaluation, supplemented with traffic and conversions from the assistant channel. Anyone who promises an exact ranking in AI answers, like a position in search, is measuring something that does not exist. The eighth myth is that this is a separate discipline with its own team and budget. In practice, most measures overlap with work that should have been done anyway: technical accessibility of the website, clear texts with specific data, order in company data across registers and catalogues, and building mentions and professional reputation. What differs is the emphasis and the way of measuring, not the substance. A company that has these fundamentals in order starts with a significant advantage, and a company that does not will not replace them with any specialised tool. The ninth myth is that it pays to block all AI robots and protect your content. The decision is legitimate, but it comes at a price: by blocking search robots you close off your route into the answers as well as the traffic that comes from them.
See also: GEO audit, Indexing delay in AI, Share of voice in AI responses.
