SEO Recap covering July 1, 2026

6 min read·7 stories

Daily summary of what matters in SEO, GEO, AEO, and AI search generated with Claude Code (beware of hallucinations)

Generative Engine Optimization (GEO) & Answer Engine Optimization (AEO)

Apple's new Siri AI (announced at WWDC June 8) reads web content to compose answers and acts inside apps, but does not navigate websites to complete tasks; it ships this fall on a Gemini-derived model.
Google's Chrome Auto-Browse arrives on Android at the operating-system level in late June, on by default with system-level authority, and drives the browser to fill forms, book appointments, and shop.
The author notes both agents run on Gemini, with Google supplying the model under both front doors.
Recommended actions: run your highest-value task flows (checkout, booking, lead capture) against an agent yourself without visual cues to find where they break, and keep server-rendered, well-structured content so reading agents can surface you in answers.

The updated randomized study by Saharsh Agarwal and Ananya Sen shows a 39.8% drop in organic clicks when AI Overviews are shown, revised up from the 38% reported in April.
Researchers found no statistically significant difference in click quality between conditions: about 40% of same-tab clicks returned to results in both groups, roughly 18% of visits ended within 10 seconds, and time on site was indistinguishable.
The findings contradict Google VP Liz Reid's claim that AI Overviews mainly eliminate low-engagement "bounce clicks," which Google has not backed with data.
Losses concentrated in informational queries, where AI Overviews triggered on 53% of searches versus 15% navigational and 6% transactional; removing a top-of-page Overview nearly doubled clicks to position one.
The paper remains a working SSRN draft that has not completed peer review, and the authors warn aggregate losses could grow as AI Overviews expand.

Lily Ray spotted AI Overview citations displaying markdown tables directly in snippets and asked whether SEOs should read anything into it.
Google's John Mueller called the behavior "unexpected" and asked for details, saying it does not mean Google treats markdown files differently from any other content page.
Barry Schwartz reminds readers Google Search does not use markdown files or llms.txt, so this is likely Google picking up a markdown file like any other file on the web.
Schwartz was able to replicate the behavior himself.

Fabrice Canel, Principal Product Manager at Microsoft Bing, announced his retirement effective July 1 after nearly 30 years, taking Microsoft's Voluntary Retirement Program.
Canel led Bing's crawling and indexing team, was closely tied to Bing Webmaster Tools, and became the public face of IndexNow, the open protocol Microsoft and Yandex launched in 2021.
Microsoft has not named a successor for Bing's crawling and indexing team or its webmaster communications; IndexNow and Bing Webmaster Tools remain available as usual.
No changes to how Bing crawls or indexes sites were announced, but SEO professionals lose a familiar point of contact on the Bing side.

Notion's "Great Renovation" report surveyed more than 6,100 AI decision-makers and users across 10 markets, finding 57% at Level 1, 31% at Level 2, 10% at Level 3, and 2% at Level 4 of its maturity model.
Senior leaders, including CEOs, operate at Levels 3 and 4 at six times the rate of individual contributors, contradicting the idea that executives push AI without using it.
Advanced (Level 3-4) organizations lead on integration (55% vs 37%), governance frameworks (42% vs 26%), and real impact metrics (37% vs 22%), while self-reported time savings decline as a measurement approach.
The gap between investment and readiness widens with maturity, from 48% at Level 1 to 68% at Level 4; Singapore leads at 21% of orgs at Level 3-4, with the US at 11%.
The author recommends mapping actual workflows against the four-level model and replacing "time saved" self-reports with one quality metric and one workflow metric.

The author cites Forrester's 2026 B2B Predictions estimating ungoverned generative AI use costs $10 billion in enterprise value, and Jasper's data showing only 41% of marketers can prove AI ROI in 2026, down from 49%.
The piece defines the "cognitive mirage" as plausible-but-untrue AI output that passes surface-level review, mapping it to Anthropic's research on confabulation in large language models.
The four-step protocol: isolate and restate the conclusion, run inverse-premise and third-party-critic devil's advocate prompts, do a human-led plus fresh-AI-chat peer review, and log hallucinations in a shared changelog.
Two B2B examples show the failure mode: AI mis-mapped intent signals to the wrong product category, and AI synthesized competitor and analyst language while passing it off as buyer voice.
The author recommends hard-coding devil's advocate prompts into workflows, routing low-confidence outputs to human review, and treating every AI output as a hypothesis until verified.

Real content from original publishers, including Press Gazette (twice), Search Engine Land, and Moz, has been removed from Google Search via fraudulent DMCA requests Google complied with.
Pedro Dias says Google does not verify submitter ID or credibility; each fake takedown can keep a URL out of results for at least two weeks, and compounded claims can stretch to months.
Dias warns that sites relying only on Search Console warnings may miss roughly 80% of DMCAs filed against them because GSC misses many of them.
Google sued companies weaponizing the DMCA route in 2023, but the industry reports the problem has grown, with some claims filed from implausible locations near Antarctica.
Charles Floate reports an engineer told him Google knows about the issue but says it can't stop it.

Published on July 2, 2026