Google's AI Overviews now cite roughly half as many sources per answer as they did in the spring. SE Ranking data shows the average fell from a peak of 16.31 sources in April 2026 to 7.66 in October 2026 — close to the 6.82 recorded back in November 2024.
The numbers come from SE Ranking, reported on November 12, 2026, and they trace a clear arc. In November 2024, an AI Overview answer cited an average of 6.82 sources. That number climbed through the winter and spring until it reached 16.31 in April 2026 — more than double where it started. Then it turned. By October 2026 the average was down to 7.66, a fall of more than half from the April peak and almost back to the level of two years earlier.
The pattern matters because the source count is, in effect, the size of the door into AI Overviews. When an answer cites 16 sources, sixteen sites get a link, a mention, and a chance of a click. When it cites fewer than eight, the same answer has room for half as many. The total number of citation slots across Google has therefore shrunk sharply since the spring, even if the number of queries showing an AI Overview has stayed the same or grown.
It is also a reminder that AI Overviews are not a fixed product. The format has changed repeatedly since launch — in length, in layout, in how sources are displayed, and now in how many sources are shown. A strategy built on the April version of the format, when answers were long and heavily sourced, would be measuring itself against a product that no longer exists in that form.
The data itself does not say why Google cut the number of sources, and any explanation should be treated as interpretation rather than fact. One plausible reading is that Google is tightening answers: fewer, more confident sources rather than a long list of supporting links. Another is that the system has become better at picking the strongest sources and no longer needs to hedge with a dozen backups. A third is that the change is about the page itself — shorter answers, cleaner layouts, and less clutter above the organic results.
Whatever the cause, the direction is consistent with a broader trend in AI search this year: answers are getting more selective. Other studies in 2026 have found citation counts falling or concentrating on fewer domains across several AI systems, not just Google. The spring peak of 16.31 may turn out to have been the outlier — a period when the format was at its most generous — rather than the baseline everyone should have planned around.
For SEOs, a smaller citation pool means fiercer competition for each slot. When sixteen sources were cited, a solid page on a relevant topic had a reasonable chance of being included. With seven or eight slots, the bar is higher: your page has to be one of the best handful of sources for that query, not merely a good one. Pages that used to appear in AI Overviews may have dropped out without any change to their own content, simply because the format now has less room.
That raises the value of being the single best source for a specific question. Broad, me-too content that could once ride along as source number fourteen is less likely to make the cut now. Original data, first-hand experience, a definitive answer to a narrow question — the things that make a page the obvious choice rather than a plausible one — matter more when the list is short. If you track AI Overview visibility, watch not just whether you appear, but how many sources the answers in your niche now carry, and whether your pages are surviving the squeeze.
There is a reporting implication, too. If you have been telling clients that AI Overview citations are growing, check the trend in your own tracked queries before repeating it. The SE Ranking figures show the average source count falling for six months. Your niche may differ — some topics may still carry long source lists — but the overall direction is down, and a report that assumes the spring numbers are still current will be wrong.
The number to watch is whether 7.66 is a floor or a waypoint. If the average stabilises near its November 2024 level, AI Overviews will have returned to roughly where they started, and the spring peak will look like a brief experiment with longer answers. If it keeps falling, the format is heading toward a small handful of sources per answer, and the competition for each slot will get harder still.
Watch, too, whether fewer sources means more clicks per source. A shorter source list could concentrate attention on the sites that remain, which would partly offset the loss of slots. That is a question only click data can answer, and it is worth tracking in your own analytics: if your pages are still cited, are they getting more traffic per citation than they did in the spring? The answer will tell you whether the shrinking door is a problem, or a filter that rewards the sites left standing.
Sources: SE Ranking data, reported November 12, 2026.
Related reading on this site: SEJ Pulse: Core Update, ChatGPT Business Hub and AI Mode's New Model, Google AI Mode Ads: US Multiturn Live, UK Still Waiting in Q4, Black Friday SEO Starts Eight Weeks Out, and November Is Already Late, AI Overview Citations: What 15 Studies Agree On, and Don't, 84% of AI Citations Point to Third-Party Sites, Muck Rack Finds, and Wikipedia Leads ChatGPT Citations at 10.1% in Ahrefs' November Ranking.
Linkbaits helps you create content that earns citations and backlinks. Start with the toolkit and build your first linkable asset this week.
Start free trial