Obility Editorial · · 4 min read
Most sudden drops are not caused by your pages
A brand that showed up in AI answers last month and barely appears this month usually prompts the same reaction: rewrite the pages. That is the last thing to try, not the first. Between one month and the next, at least four things can change. Your measurement can change. The engine can switch models or ranking. The engine can lose access to your site. And the outside pages that used to name you can change. Only after those are ruled out does your own content become the likely cause.
Working in that order matters for a second reason. If you rewrite pages in the same week an engine changed, you lose the ability to tell which change moved the number.
First confirm the drop is bigger than the noise
AI answers vary from run to run, so a drop seen in one or two checked answers is not yet a drop. Compare rates across a group of related prompts on the same engine, with the same wording, location and account state in both periods. Then check your own side of the measurement: an edited prompt, a new market or a tracker that started logging in can all look like a change in visibility.
As an illustration, a brand named in 18 of 40 answers to a group of comparison prompts in early May, and in 6 of 40 in late May, has moved from roughly 45 percent to 15 percent. The 95 percent ranges around those rates do not overlap, so the change is real enough to investigate. A move from 18 to 15 of 40 would not be.
The engines publish the dates they change models
Model changes are announced, and they are worth keeping on a timeline next to your chart. OpenAI’s help center records that on 13 February 2026 ChatGPT retired GPT-4o, GPT-4.1, GPT-4.1 mini, o4-mini and GPT-5, and on 11 March 2026 retired the GPT-5.1 models, moving conversations to GPT-5.3 Instant and GPT-5.4 Thinking. Google announced on 27 January 2026 that Gemini 3 became the default model for AI Overviews globally, and on 19 May 2026, at I/O, that Gemini 3.5 Flash became the default in AI Mode.
A new model can weigh sources differently and name a different set of brands for the same question. Google’s own documentation also states that AI Mode and AI Overviews may use different models and techniques, so a change in one need not show up in the other. Check each engine separately.
The strongest clue is what happened to everyone else. If the whole shortlist in your category reshuffled on the same date, the engine changed. If competitors held steady and only you fell, look at your own side.
Google’s AI answers sit on top of Google’s index
Google’s documentation on AI features says a page must be indexed and eligible to appear in Search with a snippet before it can be shown as a supporting link in AI Overviews or AI Mode, and that the same SEO fundamentals apply. So a ranking update can plausibly change which pages those features cite, although Google does not publish by how much.
Google’s Search Status Dashboard dates those updates. The March 2026 core update began on 27 March. The May 2026 core update began on 21 May and took almost 12 days to roll out. The September 2026 spam update began on 24 September and ran for more than 13 days. Notice the May overlap: a drop in AI Mode that week coincides with both a model switch and a core update, and dates alone cannot separate them. Search Console helps here. If the same pages lost organic clicks and positions in the same window, treat it as a search ranking problem first.
Then rule out access and the sources that named you
Access failures are the cheapest to fix and the easiest to miss. Google’s documentation lists crawling allowed in robots.txt and by any CDN or hosting infrastructure among its basics. A new bot protection setting or firewall rule can quietly block the crawlers an engine depends on. Check your server logs for the crawlers each engine documents, read by user agent, and treat a sudden fall in those requests as a signal worth checking rather than proof.
Sources are the cause teams most often overlook. If an answer used to cite three industry lists that included you, and now cites two comparison pages that do not, the engine lost the pages that carried your name. That is why each observation needs its cited sources and recommended competitors saved alongside the answer text. Obility’s Answer Engine Insights keeps the surrounding answer, the competitors it recommends and the sources it cites with every observation, so the before and after can be compared source by source. Agent Analytics keeps crawler requests separate from referral visits, so an access problem does not get mistaken for a content one.
A diagnosis narrows the cause, it does not prove it
No engine will confirm why your brand was dropped. What the order above gives you is the most likely cause and a matching response. For an engine change, keep collecting for two or three weeks and set a new baseline before you edit anything. For an access problem, fix it today. For a lost source, work on getting accurately represented in the pages the engine now cites. Only when the engine, access and sources all held steady is it time to rework your own pages, and then change one thing at a time and record the date you did.