Let’s pick up where we left off in the previous article and begin to debunk, one by one, the 10 myths we’re being fed about positioning in relation to generative artificial intelligence.
The 10 Myths about Positioning
Myth 1: “You need a special technique, different from SEO, to appear in AI results”
Put like that, this statement is not backed up by anything. Google explicitly states that existing SEO best practices remain relevant for AI Overviews and AI Mode and that no special optimisations are required. OpenAI primarily asks for proper access via crawlers and well-configured controls. Microsoft adds a focus on grounding (which we might loosely translate as ‘reliability’), citations, clarity and measurement.
Myth 2: “To display the content, you need special AI markup, a dedicated file or some mysterious technical trick”
Here too: no. Google clearly states that no new machine-readable files, AI text files or special markup are required, and that there is no specific schema.org schema to add in order to appear in its AI features. OpenAI mentions robots.txt, access for OAI-SearchBot, noindex and identifiable IP addresses. The only instance where more specific mechanisms appear is in certain verticals, such as the shopping feature in ChatGPT; however, these cannot be treated as a general rule for the entire web.
Myth 3: “Simply being mentioned frequently is enough for a brand to automatically become a reliable recommendation”
No. Both Google and OpenAI describe systems that do not function like a static SERP. Google explains that AI Overviews and AI Mode can perform multiple related searches on different subtopics and sources, and may also use different models and techniques, meaning that the set of answers and links may vary. OpenAI, for its part, says that ChatGPT Search can rephrase the query, send it to different providers, perform additional queries and cross-reference information. So, as things stand, we can safely say that an occasional appearance does not demonstrate a stably ‘secured’ position. Perhaps in the future things will change and the concept of an ‘authoritative source’ will emerge. But I don’t have a crystal ball, so I’m just reporting what we know as of today.
Myth 4: “Directories, forums, external links: you have to keep an eye on everything”
The official sources consulted do not indicate a general obligation to maintain a presence on directories, forums and every possible platform. What emerges is more specific: Google recommends keeping information such as Merchant Center and Business Profile up to date, whilst Bing highlights Bing Places for Business as useful for local listings. Full stop.
May I, however, add a personal observation of mine, backed up by absolutely nothing? It seems to me that certain communities (especially Reddit) are very often used as sources by various AIs. So who knows, perhaps things are changing.
Myth 5: “E-E-A-T is a score, or a direct ranking factor”
No. Google states this explicitly: E-E-A-T is not, in itself, a specific ranking factor. It is a framework – or, if you prefer, a checklist – that helps to identify the characteristics typically associated with the most reliable and useful content. Furthermore, Google specifies that trust is the most important factor among those associated with E-E-A-T. So anyone who talks about ‘optimising the E-E-A-T score’ as if it were an official, measurable metric is misrepresenting Google’s documentation.
Myth 6: “Google penalises content created using AI”
This is the biggest load of rubbish doing the rounds. Because, obviously, it gets clicks. And why does it get clicks? Because it creates a ‘pain point’, triggering panic among those who use AI to generate content (i.e. 90% of the web). And I don’t need to explain to you how clickbait works, do I?
Google says that the main criterion is quality, not the method of production. Don’t believe me? Then read it for yourselves on the official website. However, it also adds that the use of AI or automation to generate large numbers of pages with no added value may violate the policy on scaled content abuse. It’s a huge difference, and one that is often grossly oversimplified in debates. Here too, allow me to offer a personal observation: there are so many articles circulating on this topic that it’s only natural that, after a while, people start to believe it. But the fact is that there are no official sources on the matter. Let’s not confuse the quantity of information with its quality: it’s not true that repeating the same lie 100 times makes it true!
Myth 7: “With AI, all you need to do is stuff the text with keywords or publish lots of variations”
It feels as though we’ve gone back to those crude black-hat techniques we used to talk about 10–15 years ago. In the paper that formalised GEO, the researchers clearly state that practices such as keyword stuffing perform poorly, whilst strategies such as adding citations, relevant quotes and statistics improve visibility. As for Google, anyone working in this field has known for years that any content created to manipulate rankings is subject to penalties. So, once again, what are we talking about?
Myth 8: “AI traffic or the impact of AI cannot be tracked”
Put simply, that’s not true. Let’s be honest: there aren’t any neat, clear dashboards that tell us everything we want to know about how AI treats our brand and our company. But it’s not true that it’s all a huge mystery either!
Google includes AI-generated traffic in the overall traffic metrics of Search Console and provides some information in the Referrals section of Google Analytics. OpenAI adds `utm_source=chatgpt.com` to ChatGPT Search referrals. Microsoft has introduced AI Performance in Bing Webmaster Tools to show citations, grounding queries and referenced pages in AI responses. As things stand, therefore, measurement exists, but it is fragmented, incomplete and varies from platform to platform.
Myth 9: “If a website is blocked from crawlers, it disappears completely from search engines”
This isn’t always true either. OpenAI tells us that a page blocked from OAI-SearchBot may still appear as just a link and title if the URL comes from third-party search providers or other pages with an active crawler, and if the system deems it relevant. To prevent this as well, the documentation recommends noindex. Just as you would do for Google. In short: blocking a bot and preventing any exposure are not the same thing.
Myth 10: “AI visibility = ranking”
This is a complicated matter. Microsoft is very clear and states in black and white that AI Performance metrics do not indicate ranking, authority or the page’s role in the individual response. The GEO paper, for its part, explains that visibility in generative search engines is a different concept from the classic average position: factors such as the length of the citation, the influence of the source in the response, subjective relevance, click-through probability and other dimensions come into play. And then there is Google: Google does not comment openly on the concept of ranking, and so, in the absence of primary sources, I shall refrain from commenting. But in general, we can say that reducing everything to traditional ranking is convenient, but wrong.
What remains, once the fluff has been stripped away
Shall we try to draw up a checklist of what we know so far? OK. For Google, what matters most is content that is useful, original, reliable, technically accessible and well-presented (basically the same guidelines Google gives us for SEO). For OpenAI, what counts is proper crawler accessibility, well-configured controls, relevance and reliability, as well as specific cases such as product feeds for shopping. For Microsoft, what matters more and more is how content can be cited, verified, interpreted and reused within AI responses, with a strong emphasis on clarity, structure, evidence and up-to-date information.
So if you’re looking for some guidance, here’s what I can tell you: make your site accessible to crawlers and ensure it can be indexed; publish valuable content, not rewrites of existing material; make it clear who the author is, how the content was produced, and why it should be considered reliable; use structured data that matches the visible content; keep your business and product data up to date; configure robots.txt, noindex, preview controls and bot permissions correctly; and measure what you can using the tools actually provided by the platforms.
The conclusion
It seems to me that what’s difficult these days isn’t understanding how AI behaves, but distinguishing genuine information from sensationalism, nonsense, clickbait and those who promise ‘secret tricks’ for becoming ChatGPT’s favourite website.
I get it: it’s a new subject, everyone’s experimenting, information sources are scattered, and it’s often a nightmare sifting through verifiable facts and the rubbish written just to get clicks.
However, if you want to understand how this world works, you need to be sceptical. Check your sources, make sure that what the influencer or guru of the moment has told you is based on fact, and go straight to the primary sources. Don’t trust anyone, especially those who try to sell you secret formulas.
And with that, I bid you farewell.
See you soon.
