Ahrefs data suggests ChatGPT is citing the origin of information more often than the sites that summarize it, while Nozzle says Google may be handing rank trackers altered results. Both stories come from vendors describing their own data, so each comes with a verification step for your next report.
Who Gets Cited: Origins Gain, Explainers Slip, Shorts Get a Question Box
Ahrefs ranked ChatGPT’s most-cited U.S. domains in March and September, and the movement favors the source over the summary. YouTube, meanwhile, is preparing to let some viewers question a Short as it plays.
- Ahrefs Data: ChatGPT Citations Shift Toward Primary Sources Ahrefs says PubMed climbed to fourth from 64th while WebMD dropped to 79th from 18th, and that Reddit and Wikipedia lost large shares of U.S. answers. It is a vendor’s sample, not a statement from OpenAI, but the advice is concrete: explainer pages now compete with the sources they summarize.
- YouTube Plans a Test of Ask YouTube Inside the Shorts Player Search Engine Journal reports that a limited experiment for English-speaking viewers 18 and older would add a sparkle icon to select Shorts for asking questions about the clip. The test has not started, and a question asked inside a Short never reaches the YouTube search box where teams usually look for demand.
Measurement Trust: When the Data Looks Complete but May Not Be
A rank-tracking vendor says the problem has moved from missing data to wrong data. The claim is unreplicated, so it works best as a reason to cross-check.
- Nozzle Says Google Is Serving Rank Trackers Swapped-Out SERPs Nozzle’s Derek Perkins says 8.0% of the SERPs his firm collected were poisoned, with forums and social sites filling gaps once held by tracked sites. No independent party has replicated the figure, so treat tracker movement as unconfirmed until Search Console or a manual spot-check agrees.
Today’s Quick Hits
- Microsoft tests Decision-1 scoring model in Copilot and Bing Microsoft staff confirm only experimentation, and nothing published explains how the model figures in Bing search.
- AI Overviews will answer gibberish and no-result queries Barry Schwartz showed an AI Overview replying to keyboard mash, so an Overview appearing is not proof a query has real demand.
- Merchant Center no longer blocks products over poor images Google’s rewritten help page now says low-quality photos keep products eligible for ads and free listings, though click-through may still suffer by Schwartz’s reading.
- Google Ads tests describing a campaign goal in plain language Google confirms the typed-goal builder is only an experiment, so agencies have time to decide how they would review AI-suggested goals and assets before anything launches.
- Google Ads beta drops Target CPM for Video Reach campaigns Teams with firm cost ceilings may prefer to keep manual Target CPM, and a manual campaign run alongside the automated one gives a baseline to compare CPMs, reach, and frequency.
- Demand Gen Lookalike segments to default to suggestions Google plans to treat Lookalike segments as hints instead of hard limits, and advertisers can opt out to keep the strict targeting constraint.