(A preliminary note: We are focusing here on ChatGPT from OpenAI, as it is currently probably the most frequently used LLM (Large Language Model) in AI. Therefore, the statements might need to be relativized for other LLMs. However, it is assumed that these do not perform better or worse in terms of SEO than ChatGPT.

Let's return to the question: Can ChatGPT do search engine optimization? The short answer is: No. Not properly, at least. And not without human SEO expertise behind it.

•    Text drafts often remain superficial, wooden, or clichéd and are therefore - without further editing - usually not considered particularly relevant from an SEO perspective.
•    Keyword ideas are generated from the context of the respective topic and do not access actual search data (search volumes) to determine the actual relevance of these keywords. Therefore, they are also usually not immediately SEO-suitable.
•    Recommended text lengths: Body texts, meta titles, and meta descriptions are often much too long without further specification, often do not automatically use the most relevant keywords, and therefore often do not even meet the minimum requirements for "good" meta information.
•    There is no check for possible copyright violations, which poses a risk.
•    Nor is it always ensured that the AI does not incorporate errors, whether factual or linguistic.
•    This is particularly true for translations of the original text.

Our experience shows that the original versions of a text generated by ChatGPT (be it the text itself or even just the metadata) almost always completely fail when reviewed by specialized SEO platforms.

•    Proceed iteratively: Do not compose generic "all-in-one" prompts with the motto: "Write me an SEO-optimized blog post on topic X including metadata and headings, taking into account all keywords relevant to the topic."
•    Instead, proceed step-by-step from the general to the specific: first generate ideas and initial text drafts (if necessary, also first keyword candidates), have these edited by experienced people/copywriters; then have drafts for metadata created, check these in terms of character count and relevant keywords on a solid data basis and revise; finally, perform fine-tuning with subheadings and/or bolding.
•    At the end, take care of possible internal links, backlinks, etc.
•    Do not forget a final plagiarism check!
•    And most importantly: At and after each step, check the (intermediate) results using a "real" SEO tool such as Google Keyword Planner, Semrush, Ahrefs, Moz, etc., and adjust and further optimize if necessary!

AI can be very helpful in SEO work and can significantly speed up the process, especially when it comes to generating text ideas, writing initial drafts, summarizing a topic, or analyzing existing websites in terms of content and technical aspects. However, it does not replace the precise review of (intermediate) results based on specific and relevant (SEO) data, such as the average monthly search volume for a specific keyword; nor does it replace further editing and fine-tuning by "knowledgeable" entities - that is, real human intelligence with the corresponding SEO expertise - in or after each process step.

Therefore, for good, relevant, sustainable SEO work: NEVER (never ever) rely on AI alone! This also applies particularly to translations of supposedly "finished" SEO texts! Instead, rely on real (human) SEO expertise (or build it) and involve the experts closely in the work. Only in this way can actual relevance and thus visibility in the most important search engines be achieved.

PS: This text was written entirely without the help of AI. Honestly. 😉