Original content is becoming less common on the internet
When generation costs next to nothing, it is cheaper to rewrite someone else’s article than to conduct original research. As a result, for any query, there are a dozen texts that endlessly paraphrase one another. There is no new knowledge in them.
My position is simple: all content on a website should be original. Articles, glossary terms, FAQs, and introductory copy on service pages. Not because AI models cannot write, but because they cannot work on a project, catch an indexing bug, and explain how it was resolved. Yandex says the same in its Webmaster Help documentation: AI models can assist the author, but must not replace them.
What search engines actually check
There are many myths here, so I rely on primary sources.
Yandex outlined the framework in its help documentationEPOS: expertise, usefulness, originality, and substance. These same aspects affect both rankings in Search and inclusion in Alice AI answers. Expertise comes first and means the author’s verified competence and the website’s specialization.
Google’s logic is different in form, but the same in substance. Since 2023, the company has repeatedly said that it evaluates content quality, not how the content was created. Penalties are applied forscaled content abuse: mass production of pages to manipulate search results, regardless of who wrote them. The March 2026 update removed precisely these kinds of projects from search results.
A separate note on detection. Yandex has a public AI detector, but Yandex has never stated anywhere that its result is linked to rankings. So I would not use the claim “the search engine detects AI-generated content and lowers the website’s ranking.” The problem is not detection, but that recycled text fails to meet the standards of originality and substance–and that is enough.
How a search engine reads an author page
The mechanics look like this. A crawler opens the article, reads the structured data and heading structure, and sees the listed author. It then follows the link to that person’s page. There, it either finds a real specialist with a biography, social media profiles, and contact details, or it finds nothing.
In the second case, the byline remains a declaration. In the first, it becomes a verifiable fact: a specific person wrote the material, they have a presence beyond your website, and they stand behind what they wrote under their own name. A website’s expertise for AI models is built from details like these, not from the phrase “a team of professionals with 10 years of experience” on the homepage.
Technically, the connection is built throughschema.org: Article links to Person, Person is linked to Organization, and the entities are connected through @id. Separate structured data blocks where the article knows nothing about the author and the author knows nothing about the company do not solve this problem.
The main mistake on employee pages
I regularly see the same pattern. An author page exists, but it contains only the person’s surname, first name, and a line about which section they are responsible for. Who they are, what they did before, and how to contact them–nothing.
This is not only about signals for search engines; it is about the reader. Someone has a question about the article. There are no comments on the website. There is nowhere to write to the author. Through the contact form, the reader reaches a manager who did not write the article and cannot provide a substantive answer. The dialogue ends where it could have begun.
What a page needs in order to work:
- a real biography and experience, not just a job title in one line
- links to social media and professional profiles
- direct contact details for reaching this particular person
- author's list of materials withinternal linking to them
- specialization: which topics they cover
What we saw on our project
On the Foton website, each article listed an author with a dedicated page. To be honest, we did not separately measure exactly how this affectedAI search results , we did not. It is impossible to separate the contribution of author pages, technical fixes, and regular content publishing after the fact, and I do not want to present correlation as causation.
What we know for sure: the website activelyappears in both Yandex and Google AI search results. The author pages were part of the overall work, together with a technical audit,semantic core and systematic publication of materials month after month.
Where to start
If the website does not have author pages yet, I would proceed in the following order. Start with one complete page for a key specialist rather than five empty profiles created just to tick a box. Then add consistent markup connecting the articles, author, and organization. Next, add the author's byline to all new materials and gradually add it to older ones.
And the main thing, without which everything else is meaningless: the byline must belong to a person who actually wrote the text and can stand behind every fact in it. This is easy to check – ask the author a question about their own article.
Frequently asked questions
Not for using an AI model itself. Penalties are imposed for mass publication of pages that provide no value to readers, regardless of who wrote them.
No. A byline without a page and markup remains just words. A search engine needs an address it can follow to verify who this person is.
That is fine. One well-developed author page with a specialization and contact details works better than a set of cards with fabricated profiles.
