AI agents that will do SEO for you alone: does it really work?
On paper, AI agents for SEO seem like a perfect solution to replace a professional. However, in practice, full automation carries serious risks for indexing and reputation, while quality promotion still requires human oversight.
Author: Pavel Israelski
On paper, it sounds perfect: you connect an AI agent to Search Console and the content management system, and it researches, writes, changes, and publishes. After all, artificial intelligence is supposed to be smarter than the best SEO professional and cost less, right?
Well, in the acronym "SEO", the most important word is the one that is hardest to automate: Optimization. Keyword research, data analysis, report writing, and content creation are part of the job, but they are not the optimization itself. In the end, someone needs to change pages, meta tags, make changes to the code, publish content, create links, and take responsibility for what happens on the site.
Unfortunately, in the last year, I have been exposed to a new trend of companies promising an AI agent that will promote your site on Google. The problem is that after you go deep and understand how things work, you discover that it does only a small part of the SEO actions at best, or that the principles by which the agent works are not adapted to Google's quality guidelines, and as a result, sites experience a drop in rankings. In most cases, it turns out that it is a work tool that the SEO professional uses in the promotion process, so why mislead and sell it from the start as an agent that is the one doing the promotion?
Full disclosure: I manage an SEO agency and I myself integrate AI agents into work processes. I have no problem with them, on the contrary, I think AI is one of the best things that has happened in the SEO field and it helps us work more efficiently. My problem is with the marketing packaging "AI agent does SEO for you", simply because it is misleading. It is one of two things: either it is an SEO service provider that uses AI to save itself work time, or it is an AI agent that really does SEO in full automation, which puts the site at risk.
A 100% autonomous agent has an accountability problem
Suppose the agent identifies a weak page. It can recommend changing a title, adding content, building internal links, and fixing schema. The question is, will you also give it permission to implement this in practice?
One mistake due to an AI hallucination can damage indexing, conversions, and even the company's reputation. This is not a theoretical scenario; in a Fortune article, a case was described where Claude Code deleted a live environment and a database after confusing a real environment with an environment that was allowed to be deleted.
A CodeRabbit study on 470 Pull Requests found that code written with the help of AI contained about 1.7 times more problems, and the company Apiiro reported about 10 times more security problems among developers who used AI.
The same risk exists in content. Google does not oppose the use of AI, but its spam policy prohibits Scaled Content Abuse, that is, creating many pages without real value. In June 2026, a study by Google researchers was published on a system called S-CTS, which is used to identify synthetic spam based on repeating patterns and non-human publishing pace.
There are also things that an agent simply cannot invent
Quality content relies on experience, expertise, and new information. Google emphasizes the Helpful Content update, E-E-A-T metrics, and a patent called Information Gain, and makes sure to emphasize how much new information the page adds compared to what already exists.
An agent can summarize what is on the web, but it will not interview customers, check the product itself, or accumulate personal experience. Without a person who inputs this knowledge, you usually get a variation of what already exists, and that is the SEO of 2020.
Links and brand mentions are also not created at the click of a button. An Ahrefs study that checked 75,000 brands found that the main factors for brand recommendations in AI answers are mainly related to external signals, such as mentions on authoritative sites. If an agent generates mentions and links automatically, it might enter the manipulative zone that Google and UGC platforms are trying to block.
Results and marketing promises
Anthropic announced that it will start adding an invisible Watermark to content created with Claude. The problem starts when an SEO agent generates and publishes content automatically without human editing. This clashes with the purpose of such agents, which are intended to save human involvement.
One also needs to be careful of case studies and shiny promises. An increase in traffic of a few months is not proof that the method holds water over time. We have already seen in the case of ClickUp the 6 months before the fall that they experienced; traffic to their blog dropped by 97.6%, and they lost close to a million users due to problematic methods for creating AI content at scale.
Ask whoever is selling you the agent to show a long and stable timeline of growth. At least 8 to 12 months back so that there was at least one Core Update during the period.
Summary
If the agent does not actually do technical optimization, does not create content without expert involvement, and does not build links, then it does not "do SEO for you" but is a tool that helps an SEO professional work faster and better. I am in favor of an AI agent that helps a person do SEO better. But an autonomous agent that receives permission to change code without control might put the site at risk. If there is a person who checked, approved, implemented, and took responsibility, then it is not the agent that does the SEO, but the person who does SEO with the help of an agent.
Pavel Israelski is the CEO of Angora Media and co-founder of Chatoptic
