Google’s September 2026 spam update has reportedly finished rolling out. The discussion also highlighted fact-checking AI-generated content before publication and warnings against false author information. Alongside these developments in search quality, there is debate over how to evaluate SEO budgets, including spending on AI search.
Writing with AI and Taking Responsibility for What You Publish
Some have also suggested that the update may be accompanied by an increase in “manual actions”—penalties imposed by Google at its staff’s discretion. However, the scale of any increase and its causal link to the spam update have not been confirmed. It is important to distinguish the completion of the update from the reasons individual sites may have been penalized.
In addition to AI-powered spam detection, the discussion has addressed the need for people to verify the accuracy of AI-generated content and to avoid author information that misleads readers. The practical challenge here is not just how content is created, but also the review process before publication. There is no basis for interpreting this as meaning that content is penalized simply because it was created with AI.
For corporate communications, too, the budget for producing more content is separate from the budget for verifying it. Improving production efficiency alone does not establish safeguards against errors in product information or problems with author attribution.
Renaming SEO as GEO Does Not Change the Fundamentals
SEO is the practice of making information easier to find through search. By contrast, efforts to get information cited in generative AI responses and similar formats are called GEO. Although the terms differ, there is overlap in areas such as making pages accessible to search systems and organizing product information.
Google says that the fundamental requirements of traditional SEO remain relevant to its AI search features, and that no special technical optimization is required solely for those features. However, this cannot be assumed to apply to every AI service.
When considering budgets, the first step should be to identify work that overlaps. Introducing tools to measure visibility in AI while leaving outdated product information or pages that cannot be read by search systems in place will not resolve problems with the information infrastructure.
Separate Investments That Drive Growth from Those That Protect What You Have
Investment objectives can be grouped into four categories: increasing sales, protecting existing sales, maintaining the information infrastructure, and testing uncertain effects. It makes little sense to measure improvements to product pages that drive purchases and efforts to preserve search traffic during a site migration using the same metrics.
For example, migration work is intended less to generate new sales than to reduce the risk of users being unable to reach existing pages. Judging it a failure because traffic did not increase would miss the point. For product information improvements, reducing update errors and duplicate work can also count as success.
Set a Deadline for Evaluating AI Search Experiments
Before making new investments in AI search, companies need to decide what they want to learn. The approach should define specific questions to test—such as whether a company’s products are described accurately and which pages are cited—and use the results to decide whether to continue or stop. Rather than combining SEO and GEO across the board, distinguishing shared infrastructure work from experiments specific to each service can make duplicated costs easier to identify.
However, traditional SEO initiatives should not be retained unconditionally, either. Whether it means publishing more articles that do not recoup their production costs or continuing AI experiments with unclear objectives, companies need to compare these choices with the option of reallocating the budget to other uses.
The number of mentions in AI responses can serve as a measure of visibility, but by itself it does not prove a contribution to purchases. Search traffic, sales, information accuracy, and operating costs each measure different outcomes.
Comments