Google Rankings Swing, ChatGPT Mislabels Traffic: DesignRush SEO Roundup

Our analysts follow changes that affect how easily people find websites through search engines and AI tools. Brands building search programs can work with carefully checked SEO agencies to put the right plans into action.

Recent changes to online platforms have made it harder for brands to understand their own search traffic.

• People on WebmasterWorld reported big traffic changes after Google's spam update, while two rank-tracking tools showed results that differed by a factor of five.
• Links from ChatGPT lost their UTM tags after GPT-6, and Adobe's Cedric Huesler measured an increase in Direct visits on the same pages where visits from ChatGPT fell.
• Google announced on October 15 that AI Overviews and AI Mode would show more links within answers and place them higher, but the announcement did not say how many visits those links bring.

Over one week, two platforms changed how AI answers handle links. Neither change told brands how many visits came from those links.

Our Take: Does SEO Reporting Have a Data Problem?

Because ChatGPT's links were no longer tracked correctly, some visits from AI tools appeared as Direct traffic in analytics reports.

SEO expert Chris Long has also reported that Google's efforts to stop automated data collection are causing rank-tracking tools to return incorrect results.

We see both problems as unwanted side effects because neither change was aimed at SEO teams.

Companies already use these numbers to decide how to spend money and which pages to rewrite. A team can check its pages and links. But can it check whether the tools measuring them are working correctly? That is much harder.

Rank Trackers Falter as Traffic Swings

One week after Google's September spam update ended, people on WebmasterWorld reported that website traffic was changing every hour or two on October 15.

SEO news editor Barry Schwartz wrote that people were still actively discussing the possible update over the weekend. He said the update had not been confirmed.

Teams usually check rank-tracking data to confirm whether Google has released an update. But that data may not be reliable.

Nozzle, a rank-tracking platform based in Utah, reported that 8% of the search results pages it checked since mid-September were "poisoned," meaning the data was unreliable.

These pages looked normal but were missing most of the websites that usually appeared for a keyword.

How widespread the problem seems to be depends on which company collects the data. AccuRanker, a cloud-based rank-tracking platform, found a smaller share.

From October 1 to 9, about 1.5% of the keywords it sampled showed a different group of websites on the first page each day. Using the same rule, that was one-fifth of Nozzle's rate.

AccuRanker also said Google tightened its detection measures again on October 15, mainly for mobile searches. Google itself did not make a statement about this.

Schwartz guessed that Google was targeting AI bots and SerpApi.

Chris Long wrote on LinkedIn that Google appears to block automated data collectors by banning large groups of IP addresses and showing more CAPTCHA tests, which ask users to prove they are human.

He also said Google gives these tools pages that look normal but contain incorrect data. As a result, rank-tracking tools have become "a large casualty."

Until the problem is resolved, both Nozzle and AccuRanker advised teams to check Google Search Console. If a rank-tracking tool shows that a site has dropped but its impressions and clicks in Search Console remain the same, the site's performance may not have changed.

This means teams must rely on Google's own reports to check a problem that both companies blame on Google.

GPT-6 Raised Direct Links to 42% of Answers

Adobe's Cedric Huesler reported that on October 8, links from GPT-6 stopped carrying their UTM tags and referrer information. Analytics tools then counted these visits as Direct traffic, which is the category used when the source of a visit cannot be identified.

Adobe measured the effect on about 500 pages where ChatGPT brought in at least 5% of all visits.

In the sample, from October 9 to 13, visits from ChatGPT to those pages fell by 31,000 a day, while Direct visits rose by 26,000.

The cause turned out to be a software bug. Huesler later wrote that OpenAI confirmed the problem on October 14 and began rolling out a fix early on October 15.

OpenAI's release notes do not mention the missing tag, the bug, or the fix.

According to Tomek Rudzki, who works on generative engine optimization (GEO) and AI search optimization at the AI search analytics platform Peec AI, GPT-6 also changed which links ChatGPT includes in its answers.

He tested the same 1 million prompts before and after GPT-6. He compared results from October 2 to 7 with results from October 9 to 13.

In Rudzki's data, ChatGPT included a direct link to a website in 42% of its answers, up from 8% to 10%. Of those links, 54% led to a website's homepage.

These links do not include tracking tags, so visits they bring in are counted as Direct traffic.

Because the prompts stayed the same, a GEO team that claims credit for an increase in ChatGPT visits after GPT-6 was released is taking credit for a change made by the AI model.

Google Pushes More Links in AI Answers

In a blog post on October 15, Google announced that AI Overviews and AI Mode would show more links within their answers and place them higher up.

Google is testing a section at the top of an answer that highlights the one or two official pages needed to complete a task. It is also testing clickable creator names on videos in AI Mode. Clicking these names opens Search profiles.

Carousels showing recent articles are appearing in more places and languages. These carousels also highlight a user's Preferred Sources, or the websites the user has chosen to see.

Nick Fox, Google's senior vice president of Knowledge and Information, wrote that the number of people using this feature has grown sevenfold since the start of the year.

News publishers are also taking part in this effort. More than 400 publications now belong to the News AI pilot, which pays them for enhanced content that Google uses to develop AI features.

Google Search Console's AI report counts impressions, which measure how often these placements appear. DesignRush reported in June that the first version of this report did not include click data.

Google explains what it pays for and where it places links, but its announcement does not say how many visits those links bring.

So, if a link moves higher in an AI answer, who can tell a brand whether readers clicked it? For now, only the brand's own analytics tools can provide that information.

SEO Industry Insights

Glenn Gabe wrote on X that Google's test of an AI Mode sidebar on the main search results page is another step toward making AI Mode the default search experience, or moving toward a version of it.

SEO researcher Damien Andell said on X that the sidebar now appears in every search he makes. If the feature reaches more users, links in AI answers could be seen by many more people searching online.

Meanwhile, Dan Taylor, founder of 10 Spurs, wrote in a Search Engine Journal article that his prediction model puts the chance of a Google core update arriving within 30 days at 96.8%.

He expects Google to target large amounts of content made at scale and fake author profiles. Google's updated documentation already identifies these as quality problems.

Taylor says Google could once again judge the quality of entire websites, as it did with its 2011 Panda update. This means adding extra words to a page with little useful information will not fix the problem.

As a result, one useful check to make before late October is to look for pages built from the same template and author biographies that have been made up.

What SEO Teams Should Check Before Changing Strategy

Before changing budgets or rewriting pages, teams need to find out whether the numbers show a real change in search performance.

• Compare rank trackers. Check sudden ranking drops against clicks and impressions in Search Console before changing pages or blaming a Google update.
• Check ChatGPT traffic. Compare changes in ChatGPT referrals and Direct visits on the same landing pages to find traffic that may have been placed in the wrong category after the UTM tracking change.
• Keep AI visibility and traffic separate. Track AI Overview impressions, visits from referral links, and conversions separately. Links appearing more prominently do not guarantee more clicks.

Google continues to expand AI search. But, as our previous SEO roundup showed, reports still leave out important details about how people see and click links in AI Overviews.

If ChatGPT sends a reader to your homepage first, what does the homepage ask that reader to do?

These leading SEO agencies help brands compare platform reports with the visits recorded by their own websites.