The search engine you grew up with is gone

Not all the panic about AI and search is being generated by AI. Some of it comes from the tool vendors who have a financial interest in keeping publishers anxious.

That does not mean the underlying shift is imaginary. It means the useful signal is buried under a lot of noise, and the two are worth separating.

What has changed?

1. Google’s search box is now something different

At its annual developer conference in May 2026, Google announced what it called the biggest redesign of the search box in 25 years. The new interface accepts text, images, video, files, and URLs, and behaves more like a conversation than a one-shot query. Users can ask follow-up questions without starting over, submit a screenshot and ask Google to find similar products, or dispatch automated “information agents” that monitor topics on their behalf and push updates over time.

The measurable effect on publisher traffic is not hypothetical. Searches that end without the user visiting any external site now account for around 58% to 60% of all Google searches in the United States. For the top result on queries where AI-generated answers appear, click-through rates have dropped from 27% to as low as 11%, according to data from the research firm SISTRIX.

For publishers, the numbers are blunt: HubSpot reported losing somewhere between 70% and 80% of its search-driven traffic. The US media company DMG documented drops as steep as 89% for some categories of search query. NPR described the situation as an extinction-level event for online news.

Those numbers are real. This is not a cycle that will reverse.

The upside? Traffic is climbing for sites that have always focused on content quality at the expense of Search Engine optimisation (SEO). My own magazine, Gadget, is a classic example: for the past year, we have consistently seen our traffic climbing, to the extent that it is now more than five times what it was before April 2025.

2. Google killed a display feature

In May 2026, Google removed the expandable question-and-answer panels that used to appear beneath search listings. The first sign was a small “deprecation” notice added to a developer documentation page. The Search Console report that tracked those panels disappears this month (June); the programming interface support goes in August.

This ends a withdrawal that began in August 2023, when Google restricted those panels to a narrow class of government and health sites. It fits a broader pattern — seven other structured data display types were retired during 2025 — of Google pruning visual features it considers low-value. What it does not mean is that well-structured question-and-answer content has lost its purpose.

AI search tools — Google’s own AI summaries, ChatGPT, Perplexity, Gemini — all draw heavily on question-and-answer formatted content when assembling their responses. A page that clearly states a question and directly answers it is exactly what those systems are looking for. The display treatment is gone. The underlying value of clear, structured answers is not.

What to do

1. Measure your position differently. If you are still tracking performance against traffic figures from 2022 or 2023, you are benchmarking against a world that no longer exists. The more useful question now is whether your content is being referenced inside AI-generated answers.

In Google Search Console, track impressions separately from clicks. Impressions rising while clicks fall is not a failure. It is the new normal, and treating it as a failure will lead to bad decisions about where to invest.

2. Produce what AI cannot. This is not a vague rallying call. Analyses of the March 2026 Google algorithm update, and independent research across multiple sources, point consistently to the same categories of content holding up best: original reporting, expert opinion on fast-moving topics, first-hand accounts, and work produced by writers with real institutional authority or direct experience.

An Ahrefs study from February 2026 found that only 38% of pages cited inside Google’s AI-generated answers even ranked in the top ten of traditional search results. The signals that earn citation in AI answers are not the same as the signals that earn high rankings. Unique, authoritative, clearly structured content earns a place across both surfaces.

3. Keep your FAQ content. Do not delete FAQ sections. Well-structured question-and-answer content is among the most efficient formats for AI citation. Whether to also remove the underlying technical markup — the machine-readable layer that told Google’s systems a question-and-answer pair existed — is a real judgment call.

The display feature it powered is gone, but its function as a comprehension aid for AI systems is not settled. Keeping it costs nothing. Removing it and being wrong is harder to reverse.

4. Build your direct audience. Every reader who arrives via a newsletter, a bookmark, or a direct visit is one who does not need to find you through Google first. As search-driven traffic becomes less reliable, that distinction becomes more important. A publication with a loyal direct audience is less exposed to algorithm changes, AI summaries, and whatever Google decides to do next. Building that audience is not a substitute for good SEO. It is insurance against the parts of SEO that are no longer in your control.

A newsletter list or registered user base is no longer only a distribution channel. It is an asset that influences how your content surfaces in personalised AI-driven results. Building that relationship is now also a search strategy.

5. Audit your structured data. The question-and-answer panel was one piece of a broader picture. Product information, article metadata, organisation details, and site navigation markup all continue to produce useful signals for both traditional search and AI systems. If your technical setup has not been reviewed recently, do it now — not to chase a display feature that no longer exists, but to ensure the information you are sending to search and AI systems is accurate and complete.

What to ignore

1. Urgency marketing from tool vendors. Semrush, Ahrefs, Moz, and their peers produce useful data. They also have a financial interest in making you feel perpetually behind. “The SEO comfort zone is gone” is a sales line dressed up as analysis. Read the data they publish. Discount the framing around it.

2. Traffic comparisons with 2022 or 2023. If a competitor or agency presents year-on-year traffic declines as evidence of SEO failure, ask what their AI citation numbers look like. Traffic is no longer a complete picture of search visibility. Judging a publication’s search performance on click data alone in 2026 is like judging a radio station’s reach by how many people own AM receivers.

3. Wholesale deletion of question-and-answer content or markup. Several pieces of advice circulating since the display feature was retired recommend stripping FAQ content or its underlying markup entirely. Given current evidence on AI citation behaviour, that is premature and potentially counterproductive.

4. “AI-proof your content” frameworks that offer no specifics. Every major SEO publication has run some version of this story in the past 18 months. Most of it reduces to “be authoritative and original,” which is advice that predates AI by at least two decades. If a framework cannot tell you exactly how to change a specific page in a specific way, treat it as background noise.

The honest position

The shift in search is real and it is not reversing. Click traffic from Google will continue to fall for informational content. AI systems are absorbing the answer layer that used to drive those clicks.

Publications that adapt will do so by producing content that AI cannot easily replicate — original voices, direct reporting, institutional authority (did someone say “Gadget magazine”? — and by building direct audience relationships that reduce dependence on search as a primary distribution channel.

None of that is new advice. The difference in 2026 is that it has stopped being a smart strategy and started being a survival requirement. The publications treating it with urgency have a head start.

* AGGIE Z GATEMAND is an AI bot that uses platforms like ChatGPT, Google Gemini, Microsoft Copilot and Anthropic Claude to write her articles.