SEO, AI and ChatGPT: are there news on the horizon?

Artificial intelligence and SEO: will experts really be replaced by AI?

A detailed article from Search Engine Land highlights a crucial theme of our times, particularly in digital marketing: can AI effectively perform technical SEO analysis using raw crawl data?

The research examines an in-depth test conducted to answer this question, providing insights into how GPT and AI, in general, can tackle this task

The results? Perhaps, for now, we can feel somewhat relieved: technology still requires human intervention to adequately perform and respond to tasks like this.

But for how much longer?

The potential of AI in technical SEO

The research from Search Engine Land emphasizes that many marketing professionals may find it challenging to use technical tools like OpenAI APIs, Python and spreadsheets to obtain detailed and in-depth SEO insights.

These tools, though powerful, require advanced knowledge of programming and data analysis techniques, skills not possessed by all marketers.

However, ChatGPT's user-friendly interface represents a significant breakthrough in this context.

Thanks to its ease of use and ability to interpret and respond to complex questions in natural language, ChatGPT could make AI accessible even to those without technical skills.

This means that even less experienced marketers can benefit from AI to enhance their SEO strategies without needing to learn programming or manage complex datasets.

ChatGPT Tests with Screaming Frog Data

For the test, the Butchersdogfood.co.uk website, a medium-sized UK-based dog food seller, was used. The site was fully scanned using Screaming Frog, a well-known SEO analysis tool that collects detailed data on all site elements.

The scan revealed over 3,000 internal addresses, exported into a CSV file for further analysis. This CSV contained a wide range of information, including URLs, metadata, canonical tags, response times, and file sizes.

Once obtained, the CSV file was uploaded to GPT-4o to obtain SEO recommendations.

Results and recommendations analysis

ChatGPT's responses were predominantly generic, with suggestions such as optimizing images to reduce loading times or implementing canonical tags correctly.

Despite having access to specific data from the CSV file, ChatGPT failed to provide detailed and useful advice.

For example, instead of specifically identifying which images needed compression or which pages had canonical tag issues, the recommendations remained superficial.

‍ChatGPT advised ensuring all images were optimized and that each page had a self-referencing canonical tag, without delving into specific elements requiring attention.

This level of generality makes it difficult for marketing professionals and SEO specialists to implement concrete changes based on the provided data.

The comparison with Ahrefs

The research compared ChatGPT's recommendations with the results of a technical audit conducted by Ahrefs, a leading SEO tool in the market.

Despite Butcher’s Dog Food website scoring well overall in Ahrefs' audit, similar issues to those suggested by ChatGPT were identified, such as the need to optimize redirects and image compression.

Unlike ChatGPT's generic recommendations, however, Ahrefs' audit provided more specific details, pinpointing exactly which URLs had redirect issues and which images needed compression.

The comparison highlighted that while ChatGPT can identify problematic areas in a general sense, professional SEO tools like Ahrefs excel in providing practical and specific recommendations necessary for implementing effective improvements.

Conclusions

While GPT-4o has shown improvement over previous models, it still lacks the precision required for reliable technical SEO analysis: while AI can offer some valuable insights, its recommendations often lack specificity and accuracy.

The future holds promising developments, and there is certainly potential, but currently, it's fair to say that AI cannot fully replace SEO experts.

‍

You can read the full article with the original test here.