Kevin Indig Finds SEO Losses Concentrated in Publishers and Big Sites

Similarweb estimates in Kevin Indig’s October 26 analysis show why a broad SEO decline is a weak benchmark for an individual business. Marketers need comparable websites in the same market before attributing their own result to AI.

The study’s smallest-site group had a roughly flat median result over two years, while publishers lost about 42% of estimated organic visits in both its US and French samples. The largest-site group’s combined traffic declined in both samples. Site size means a traffic quartile, one of four groups ranked by traffic. It doesn’t measure the size of the business behind the website.

Indig, who writes Growth Memo and advises companies on organic growth, says the benchmark needs to:

“fit your industry and market”

The choice of comparison sites changes which traffic decline a marketing team is being asked to explain.

Our recent coverage of Indig’s Direct-traffic analysis examined whether unidentified visits could be credited to AI. This study asks a different question: which websites belong in the comparison when judging an organic decline?

The typical site and the total tell different stories

Indig used Similarweb estimates covering September 2023 through August 2026. The organic headline compares August 2024 with August 2026. It includes 8,658 US sites and 7,730 French sites with complete organic estimates throughout the dataset. These are different country lists, not representative samples of each national web. Industry labels are provisional.

Among the small US sites in Indig’s comparison, combined organic traffic rose 19.3%, while the median rose 2.7%. Removing the 100 largest small-site gainers in each country turned combined growth negative.

The median identifies the result in the middle of the group. Adding all the visits together gives more weight to whichever sites contribute the largest changes. A few unusually strong performers can therefore make a group’s combined growth look much better than the experience of its typical member.

Keep both figures in a peer report. The typical site’s experience and the movement of total traffic answer different questions.

The same discipline applies when choosing the sites. Group a software seller with relevant software sellers before comparing it with a publishing business. Match the country and reporting period. Write down why each site belongs so the comparison doesn’t quietly change when a new report produces a more convenient average.

Check what the measurements include

Similarweb’s channel documentation defines organic search as unpaid clicks from search results and paid search as clicks from search advertisements. Its Marketing Channels reports cover web traffic, excluding app traffic. A channel’s share of visits is also a different measure from the number of visits it delivered.

Keep that distinction visible when reviewing a loss. A channel can account for a smaller portion of a site’s traffic while another channel grows. The share alone doesn’t tell the reader how many organic visits disappeared. Compare the same metric across the same period before deciding that two charts describe the same movement.

Historical versions matter too. Similarweb says its April 2026 classification update recalculated 37 months of data. That doesn’t establish a problem with Indig’s comparison. It gives teams a reason to check that older exports and current estimates use a consistent basis before joining them into a trend.

Bring the comparison back to your own pages

Google’s traffic-drop guidance recommends reviewing Search Console history for seasonal patterns and comparing similar periods. It also advises checking whether the change is concentrated in particular queries or pages, rather than relying on the site’s total.

For a software marketing team, that means separating product and documentation pages from articles aimed at broad informational searches. A loss confined to one group gives the team a narrower question to investigate. Our coverage of listicle performance shows why page format deserves attention alongside the overall traffic line.

Google also recommends examining clicks and impressions together and using Google Trends to investigate changing search demand. These checks don’t isolate an AI effect. They help locate the change before a team chooses an explanation.

Indig’s analysis doesn’t establish whether AI or paid search caused the organic losses. It measures estimated visits, not revenue. The issue is sponsored by Semrush and promotes the author’s paid resources, giving the analysis commercial context alongside its findings.

The next SEO investment discussion should start with a named peer group and a consistent measurement. Put your site’s result alongside both the median and the combined change. Show which pages are losing clicks before presenting AI as the explanation.