All issues
Discovery Digest · 26 June 2026
Issue 06. The week OpenAI went full-stack and Google cleaned house
Two infrastructure moves, a spam update with real teeth, an agentic work report that should make every content team nervous, and a quick H2 data hygiene reminder. A lot landed this week. Here is what matters and what to do about it.
01 · Search Rankings / Spam Policy
Google releases June 2026 spam update
- What
- Google has launched its June 2026 spam update, a broad rollout targeting sites that violate Google's spam policies. Spam updates recalibrate how Google's systems detect and demote manipulative, low-quality, or deceptive content across the index, and they typically produce more volatile ranking shifts than core updates because they target policy violations directly rather than quality signals broadly.
- When
- Announced via Google Search Central staff on LinkedIn, week of 23 June 2026. Rollout windows for spam updates typically run one to two weeks from announcement.
- How it shifts discovery
- If you have seen ranking drops in the last few days and your site operates in niches historically associated with spam (thin affiliate, auto-generated content, link schemes, cloaking), this update is your most likely culprit. Even clean sites can see collateral movement during rollout as Google recalibrates thresholds. Do not make reactive changes mid-rollout. Wait for the rollout to complete, then audit your site against Google's spam policies before touching anything. What I'd do: Pull your Google Search Console performance report filtered to the last 14 days and flag any queries or pages showing a drop of more than 20% in clicks. Cross-reference those pages against Google's spam policy list. If pages are clean, hold. If they are not, prioritise remediation over any other SEO work this week.
- Questions to ask
- Which of our pages, if any, rely on tactics that sit close to Google's spam policy lines, and have we audited them recently?
- Do we have a baseline snapshot of our rankings from before 23 June so we can accurately attribute any movement to this update rather than other variables?
- If we have been hit, do we know whether the issue is site-wide or isolated to specific templates or content types?
- Sources
02 · AI Infrastructure / GEO
OpenAI launches Jalapeño, its first in-house AI inference chip
- What
- OpenAI has built and launched Jalapeño, its first proprietary AI chip, designed in partnership with Broadcom and purpose-built for LLM inference workloads. This means ChatGPT, the API, Codex, and future agentic products will increasingly run on OpenAI's own silicon rather than being entirely dependent on third-party GPU supply, most notably Nvidia.
- When
- Announced by OpenAI on 24 June 2026.
- How it shifts discovery
- This is not a product you will use directly. It is an infrastructure move with downstream consequences that matter to anyone building on or competing with OpenAI's products. Custom silicon built for inference (generating outputs) rather than training means OpenAI can reduce latency, cut per-token costs, and scale capacity without being constrained by external chip supply chains. For teams doing GEO (generative engine optimisation), lower inference costs historically correlate with expanded model usage, more aggressive feature rollouts, and faster iteration on products like SearchGPT and the ChatGPT browse experience. If ChatGPT's AI search gets faster and cheaper to run, OpenAI has less reason to throttle it. What I'd do: No immediate action required on this one, but if your GEO strategy is nascent, use this as a signal to accelerate it. OpenAI is clearly investing in the infrastructure to serve AI search at scale. That is a long-term commitment, not an experiment.
- Questions to ask
- Does our GEO strategy account for ChatGPT and AI search surfaces as first-class channels, or are we still treating them as secondary to Google?
- How dependent are our content workflows on OpenAI's API, and does improved inference capacity change what we can afford to do at volume?
- What does reduced per-token cost mean for competitors building on the OpenAI API, and how does that shift the competitive landscape in our sector?
- Sources
03 · AI Agents / Content Workflows
OpenAI's agentic work data shows non-technical adoption is accelerating fast
- What
- OpenAI published an economic research paper on 25 June 2026 measuring how Codex, its agentic coding tool, has transformed work inside OpenAI and among external users. The headline numbers: by May 2026, 97.9% of OpenAI's own active users were using Codex as their primary AI tool, and Codex accounts for 99.8% of output tokens generated internally. Non-developer adoption among individual users rose 137 times since August 2025.
- When
- Published by OpenAI on 25 June 2026, covering usage data from August 2025 through June 2026.
- How it shifts discovery
- The significance here is not Codex specifically. It is the pattern. OpenAI is documenting, with internal data, how quickly agentic AI goes from an engineering novelty to the default tool across every function, including Legal, Finance, and Recruiting. Non-technical users at OpenAI now generate more than 85% of their output tokens via Codex. That adoption curve is steep and it does not flatten. For growth and SEO teams, this should reframe how you think about AI-assisted content and research workflows. The question is no longer whether your team should use agentic tools. The question is whether you are using them for tasks that genuinely take hours of human work, because that is where the leverage is. The paper also notes that nearly a quarter of all Codex requests are for tasks that would take a person more than one hour to complete. That is the usage pattern worth emulating. What I'd do: Map your team's most time-consuming recurring tasks, think keyword research at scale, content briefs, competitive analysis, GA data pulls and identify which ones could be handed to an agentic workflow. Start with one. Measure the time saved. Then expand.
- Questions to ask
- Which tasks in our search and content workflow currently take more than one hour per person and are therefore the highest-value targets for agentic automation?
- Are we using AI for single-turn interactions only, or have we started delegating longer-horizon tasks with tool access and iteration?
- If non-technical teams inside the company that built these tools adopted Codex faster than engineers once they had access, what does that tell us about the barrier to adoption for our own non-technical growth and marketing colleagues?
- Sources
04 · Measurement / Analytics
Google Analytics releases H2 2026 halftime data checklist
- What
- Google Analytics published a halftime checklist on 25 June 2026 aimed at helping teams verify their measurement setup is accurate and complete before the second half of the year. It is timed to the halfway point of 2026 and is designed as a data hygiene prompt for GA4 users.
- When
- Published by Google Analytics on 25 June 2026.
- How it shifts discovery
- The halfway point of the year is one of the best natural forcing functions to audit your analytics setup, and it is one that most teams skip. Bad data compounds. If your GA4 configuration has been drifting since January, every decision you make in H2 is built on a shaky foundation. This is particularly pointed right now given how much the measurement landscape has shifted: GA4's reporting interface has changed, consent mode requirements have tightened in several markets, and the rise of AI-driven traffic makes source attribution harder to trust without a clean setup. What I'd do: Run through Google's checklist this week. Prioritise checking your conversion event configuration, your channel grouping rules, and whether your consent mode implementation is correctly suppressing or modelling data as intended. If you have not validated your Search Console and GA4 integration recently, check that too.
- Questions to ask
- When did we last audit our GA4 conversion events, and are we confident they are firing accurately and without duplication?
- Is our consent mode implementation correctly handling users who decline tracking, and are we modelling data rather than dropping it entirely?
- Have our channel grouping rules been updated to account for newer traffic sources, including AI referrals from tools like Perplexity or ChatGPT browse?
- Sources
05 · Professional Development / Search Community
Google Search Central Live Deep Dive autumn Europe vote closes 1 July
- What
- Google Search Central has opened voting for the topic of its next Live Deep Dive event, aimed at search practitioners in Europe this autumn. The vote closes on 1 July 2026.
- When
- Vote announced week of 23 June 2026, closes 1 July 2026. The event itself is scheduled for autumn 2026.
- How it shifts discovery
- This is a direct line to influence what Google Search Central covers publicly. These Deep Dive sessions are genuinely useful: they tend to feature Search Central engineers explaining how specific systems work, which is a level of detail you rarely get from documentation alone. If your team has a standing question about how a particular ranking signal or policy is applied, voting and attending is one of the most efficient ways to get a credible answer. What I'd do: Check the vote before 1 July, pick the topic most relevant to your current challenges, and put the event in the diary now before it fills. European practitioner-focused sessions from Google Search Central are not frequent.
- Questions to ask
- Which topic on the vote shortlist maps most directly to an open question or uncertainty we have about how Google handles our site?
- Do we have team members based in Europe who should be registered to attend and bring back notes?
- Are we making enough use of Google Search Central's official resources and events as a primary source, or are we relying too heavily on third-party commentary?
- Sources
Key takeaways
What to walk away with this week
The June 2026 spam update is live. Audit your Google Search Console performance data before making any ranking changes, and wait for the rollout to complete before drawing conclusions.
OpenAI's Jalapeño chip is a long-term infrastructure signal that AI search will get faster and cheaper to run. Treat GEO as a first-class channel now, not a future consideration.
OpenAI's own internal data shows agentic AI adoption across non-technical teams is accelerating faster than engineering adoption. The teams that map their highest-effort tasks to agentic workflows first will have a structural advantage in H2.
The halfway point of 2026 is an ideal moment to audit your GA4 setup. Bad measurement compounds over a full year. Use Google's checklist as a structured starting point.
The Google Search Central autumn Europe Deep Dive vote closes 1 July. Vote for the topic your team needs most, and register early.