Organic search is not dying, but the economics of where to invest have shifted in ways that most dashboards will not show you.
Table of contents
- What this is about
- What the source says
- What the data shows
- What it does not mean
- What to do
- Numbers at a glance
- FAQ
Key takeaways
- Seer Interactive’s 2026 study found organic CTR on AI Overview queries rose from 1.3 % to 2.4 % between December 2025 and February 2026 — an 85 % gain in two months that Seer itself calls a levelling off, not a recovery.
- AI Overview presence varies more by query format than by intent category: comparison queries trigger an Overview 95.4 % of the time, while single-word queries do so only 27.3 % of the time.
- An Ahrefs study of 1,885 pages found no statistically significant uplift in AI citations from adding JSON-LD, across Google AI Overviews, AI Mode or ChatGPT. The only statistically significant result was a 4.6 % fall in AI Overview citations.
- Branded web mentions correlated 0.664 with AI Overview visibility in Ahrefs data, against 0.218 for backlinks — a gap large enough to change where link-building budgets go.
- Being cited inside an AI Overview is worth roughly 120 % more clicks per impression than going uncited on the same page, and still leaves you 38 % behind results with no Overview at all.
What this is about
Paul DeMott, writing in September 2026, lays out three priorities for 2027 SEO: allocate content by funnel stage, build structured content that machines can extract, and shift authority-building toward brand mentions and earned media. For a site owner, the practical consequence is that query format now predicts AI interception better than query intent, and the schema sprint your team just finished may have moved nothing.
What the source says
DeMott’s piece draws on several distinct datasets, and the claims are worth separating by source rather than reading them as a single argument.
On click-through rates, the piece cites Seer Interactive’s 2026 study, which found that organic CTR on AI Overview queries reached a low of 1.3 % in December 2025 and climbed to 2.4 % by February 2026. Seer describes this as a levelling off rather than a recovery and warns against drawing forecasts from two months of data. DeMott echoes that caution.
On zero-click rates, the piece draws on a SparkToro study. The U.S. zero-click baseline sits at 68.01 % as of August 2026, up from 58.5 % in the April 2025 figure. Informational queries skew above that baseline; transactional and navigational queries remain the most click-protected.
On AI Overview presence by format, Seer’s April 2026 data shows comparison queries triggering an Overview 95.4 % of the time, review queries at 86.3 %, question-format queries at 85.9 %, and single-word queries at only 27.3 %. The impressions column matters here: question-format queries carry 1.4 million impressions in the sample against 79,436 for comparison queries, so the absolute volume of displaced clicks sits with the question format, not the comparison format.
On citations and content type, Omniscient Digital found that for branded queries, earned media accounted for 48 % of citations, commercial brand content 30 %, and owned content 23 %.
On schema, Ahrefs took 1,885 pages that gained JSON-LD between August 2025 and March 2026 and set them against 4,000 matched controls. AI Overview citations fell 4.6 % — the only result Ahrefs describes as statistically significant. AI Mode rose 2.4 % and ChatGPT rose 2.2 %, both within the noise band.
A separate searchVIU experiment tested how five systems (ChatGPT, Claude, Perplexity, Gemini and Google AI Mode) handled page content during real-time retrieval. All five read visible HTML only; JSON-LD, hidden Microdata, and hidden RDFa were not read during that retrieval step.
Cyrus Shepard went at the question from the other side, ranking 23 candidate citation factors by how much evidence stands behind each one, drawn from 54 studies, experiments, patents and case write-ups. Structured data came out mid-table. Shepard’s analysis also noted that models do not appear to use schema in a schema-like way.
On brand mentions versus backlinks, Ahrefs found that branded web mentions correlated 0.664 with AI Overview visibility, against 0.218 for backlinks. In a follow-up that December, the strongest association of the lot belonged to YouTube mentions, about 0.737, and it held across AI Overviews, AI Mode and ChatGPT alike. Ahrefs is explicit that these are correlations, not proven causes.
On markup specifically for AI features, Google’s own documentation states there are no additional technical requirements for appearing in AI Overviews or AI Mode and that special schema.org structured data is not needed. Google does recommend that structured data match the visible text. Microsoft is the exception. Fabrice Canel said in March 2025 that schema helps its models make sense of a page for Copilot. That claim is narrower, and it covers one platform.
FAQ schema has been winding down for years. Google narrowed it to government and health sites in August 2023, then dropped the rich result entirely on May 7, 2026. Search Console reporting for the feature followed in June 2026, with API support ending in September 2026.
DeMott also revisits a November 2025 exchange in which someone asked John Mueller whether swapping AI text for human writing would rescue a struggling site. Mueller said he would focus on the value a site adds to the web, not on whether content was AI-generated, and that rewriting AI content by hand does not make it authentic.
What the data shows
The Seer CTR finding is the most-cited number in DeMott’s piece, so it needs the closest look.
What Seer measured: organic CTR on queries where an AI Overview appeared, tracked monthly. The two data points are December 2025 (1.3 %) and February 2026 (2.4 %). The 85 % rise is arithmetically correct. Seer warns against forecasting from two months of data, which means the figure cannot be treated as a universal baseline. DeMott acknowledges this; so does Seer.
What the format breakdown shows: the 95.4 % AI Overview rate on comparison queries is striking. But that bucket holds 79,436 impressions against 1.4 million in the question-format bucket, which runs at 85.9 %. Any strategy built to avoid AI interception by avoiding comparison content is optimising for the smaller pool. The more telling exposure is in question-format content, simply because that is where the volume sits.
What the Ahrefs schema study shows, and what it does not: Ahrefs tested pages that were already heavily cited before adding JSON-LD. The study measures whether schema bolted onto an already-visible page lifts it further. It does not measure whether schema helps an unknown page get discovered in the first place. That is a meaningful gap. The study’s design is disclosed, the sample sizes are stated, and Ahrefs is clear about the correlation-versus-causation boundary. Those are the method points I look for, and most of them are present.
What the mention-correlation data shows: a correlation of 0.664 between branded web mentions and AI Overview visibility is a moderate-to-strong association in a dataset of this type, but Ahrefs is explicit that it does not establish causation. The gap between that figure and the 0.218 backlink correlation is large enough to suggest that mention-building deserves budget parity with link acquisition — not that links have stopped working.
What it does not mean
The CTR recovery figure is the one most at risk of overreach, so I want to be direct about it.
An 85 % rise from 1.3 % to 2.4 % over two months does not mean AI Overviews have stopped suppressing clicks. A CTR of 2.4 % on a query set that previously converted at higher rates is still a suppressed number. Seer’s own framing — “levelling off” — is the accurate one. Two months of data cannot establish a trend, and DeMott says so.
The Ahrefs schema finding is being read in two incompatible directions. Some teams are treating it as proof that schema is useless for AI visibility. Others are ignoring it because the study only tested already-cited pages. Both readings miss the actual finding: schema added to a page that is already performing does not appear to increase AI citations, and during real-time retrieval, the five systems tested by searchVIU read visible HTML rather than JSON-LD. That is not the same as saying schema has no role anywhere in a pipeline that includes crawling, indexing, and grounding. The honest position is that the evidence for schema as an AI citation lever is weak and contested, not that schema is worthless across the board.
None of this survives contact with the tactics that thin out under testing, and the mention-correlation data does not mean you should stop building links. Google’s own documentation still describes PageRank as part of how it determines which pages are most helpful. The correlation data suggests that mention-building should sit alongside link acquisition in the budget, not replace it.
Finally, the funnel-stage logic does not mean abandoning informational content entirely. DeMott’s actual position is that TOFU should be deprioritised until MOFU and BOFU opportunities are covered, and that informational content still builds topical authority and internal linking structure. The “informational content is dead” reading of this data is not what the data says.
What to do
These are ordered by effort, lowest first.
1. Audit your format exposure before your intent exposure. Pull your top-traffic informational pages and sort them by query format: comparison, question, review, price, best-of, single-word — against Seer’s AI Overview rate table. A page ranking for a comparison query faces a 95.4 % chance of an Overview sitting above it. Knowing that changes how you prioritise optimisation, regardless of whether the underlying intent is commercial or informational.
2. Add citation tracking to your rank reporting. Ranking first while three competitors are cited inside the Overview is a result your current rank tracker will record as a win. It is not a win in any revenue sense, and the gap between citation volume and business impact is the part worth measuring. GA4’s AI Assistant channel (added May 2026) and Search Console’s Generative AI performance reporting (launched June 2026, impressions only) give you partial visibility. Neither is complete. GA4 excludes Perplexity and misses app traffic without a referrer, and Search Console does not yet report clicks. Running both is still better than running neither.
3. Restructure MOFU pages for extractable passages. Answer engines lift passages that stand alone. If a paragraph only works once you have read the two before it, nothing will lift it. A table row, a definition-led sentence, or a clearly labelled comparison will. Rewriting for extraction does not require a full content rebuild; it often means adding a summary table, reordering a section so the conclusion leads rather than follows, and naming entities explicitly rather than using pronouns.
4. Shift a share of your authority budget toward earned media and mentions. Most of what gets cited is not yours to publish, which is why maintenance of the pages you already own and third-party presence pull in the same direction. If your current link-building programme produces links but little brand mention volume on third-party sites, podcasts, or video, the Ahrefs correlation data gives you a reason to rebalance. The 0.664 mention correlation against 0.218 for backlinks is not a causal proof, but it is a signal that digital PR and community presence deserve a budget line alongside traditional link acquisition.
Numbers at a glance
| Signal | Figure | Source | Caveat |
|---|---|---|---|
| Organic CTR on AI Overview queries, Dec 2025 | 1.3 % | Seer Interactive, 2026 | Two-month window; sample not fully disclosed |
| Organic CTR on AI Overview queries, Feb 2026 | 2.4 % | Seer Interactive, 2026 | Seer calls this a levelling off, not a recovery |
| AI Overview presence, informational queries | 36 % | Seer Interactive, Apr 2026 | By intent classification |
| AI Overview presence, comparison format | 95.4 % | Seer Interactive, Apr 2026 | 79,436 impressions in sample |
| AI Overview presence, question format | 85.9 % | Seer Interactive, Apr 2026 | 1,408,843 impressions in sample |
| AI Overview presence, single-word queries | 27.3 % | Seer Interactive, Apr 2026 | 25,042,440 impressions in sample |
| U.S. zero-click rate, Aug 2026 | 68.01 % | SparkToro | Up from 58.5 % in Apr 2025 |
| Cited vs. uncited CTR uplift on AI Overview SERPs | ~120 % more clicks per impression | Seer Interactive, 2026 | Cited pages still trail no-Overview results by 38 % |
| Earned media share of branded AI citations | 48 % | Omniscient Digital | Owned content: 23 %; commercial brand content: 30 % |
| JSON-LD addition, AI Overview citation change | −4.6 % | Ahrefs, Aug 2025–Mar 2026 | Only statistically significant result; tested already-cited pages |
| JSON-LD addition, AI Mode citation change | +2.4 % | Ahrefs, Aug 2025–Mar 2026 | Within noise band |
| JSON-LD addition, ChatGPT citation change | +2.2 % | Ahrefs, Aug 2025–Mar 2026 | Within noise band |
| Branded mention correlation with AI Overview visibility | 0.664 | Ahrefs | Correlation, not causation |
| Backlink correlation with AI Overview visibility | 0.218 | Ahrefs | Correlation, not causation |
| YouTube mention correlation across AI Overviews, AI Mode, ChatGPT | ~0.737 | Ahrefs, Dec 2025 | Strongest single correlation in the follow-up study |
| FAQ rich results retirement date | May 7, 2026 | Search Console reporting ended June 2026 |
FAQ
Does the Ahrefs schema study mean I should remove structured data from my pages?
No. The Ahrefs study tested whether adding JSON-LD to already-cited pages increased AI citations further. It found no significant uplift and a statistically significant 4.6 % fall in AI Overview citations. That result applies to the incremental value of schema on pages with existing visibility. It does not test schema’s role in crawling, indexing or entity disambiguation. Google’s documentation still recommends that structured data match visible text. The practical position is to keep schema where it serves Google Search features that still use it (Product, Article, Organization) and where it clarifies entity relationships, but to stop presenting it internally or to clients as a proven AI citation lever.
If branded mentions correlate more strongly than backlinks with AI Overview visibility, should link building stop?
The correlation data from Ahrefs — 0.664 for branded mentions against 0.218 for backlinks — is a signal about relative association, not a causal proof about either factor. Google’s documentation continues to describe PageRank as part of how it determines which pages are most helpful for a query. The practical implication is that mention-building (digital PR, podcast appearances, video presence, third-party community footprint) deserves budget parity with link acquisition, not that links have become irrelevant. Ahrefs is explicit that these are correlations; treat them that way.
The CTR data shows an 85 % rise in two months. Does that mean AI Overviews are no longer suppressing organic traffic?
No. The rise is from 1.3 % to 2.4 %, which is arithmetically 85 % but still a suppressed CTR relative to pre-Overview baselines on the same queries. Seer Interactive describes it as a levelling off and explicitly warns against forecasting from two months of data. The U.S. zero-click rate simultaneously rose from 58.5 % in April 2025 to 68.01 % by August 2026 in SparkToro’s data, which runs in the opposite direction. Two months of CTR movement on one query set does not override the broader zero-click trend. Watch the next six months of Search Console Generative AI impressions data before drawing any directional conclusion.
The single most useful thing you can do this week takes an hour. Open your rank tracker, find your top five MOFU pages, and check whether any of them appear inside an AI Overview without being cited. If one does, your reported position is not the result you actually got.
