Loading today’s stories
Build1 distinct publisher3 min readPublished
Drawing a box around part of a web page and sending it to Gemini is a small user gesture. In a managed fleet it is a policy question, and the policy is coarse.
The Engineer · Build desk
Compiled by The EngineerSomething wrong?How this is made
The mechanism is what makes this an administrative matter rather than a feature note. The capture path runs through the browser chrome, not the page: the user opens the Gemini side panel, picks Select from screen, and draws around the region to share [2]. Whatever is inside that box becomes the basis for the next interaction [2]. Nothing about that flow requires the site to cooperate, be aware, or consent. A page that would never permit an integration, an internal admin console or a customer record view, is a rectangle like any other.
That is the part worth sitting with. Google's own framing is that the region can hold text and/or images [3], which means the sensible mental model is not "a text prompt" but "a crop of whatever is rendered". Screenshots have always been possible, of course. The difference is that the transfer to a third-party model is now one menu item inside the tool people already have open, and the source material says the value on offer is precisely the removal of that friction [4].
Meanwhile the control surface stays blunt. What the source describes for managed environments is a set of yes/no checks: is Gemini in Chrome enabled in Chrome's AI innovations settings, do account, region, subscription or rollout conditions apply, do Workspace or other administrative controls permit it, and what internal guidance should exist before staff share sensitive page content with an assistant [5]. Three of those four are switches. The fourth is a document you have to write yourself, and it is the only one that operates at the granularity the risk actually has, because there is no per-site exclusion in that list.
Availability adds a second layer of uncertainty that undercuts any attempt to standardise a process. The supplied documentation confirms the workflow but does not establish a universal price, a single worldwide availability date, or that every related feature is included with every account tier [6], and Google has said some capabilities may be limited by rollout phase, region, or Google AI Pro and Ultra subscription access [7]. Select from screen also sits inside a wider Gemini in Chrome effort that connects to Calendar, Gmail, Maps and YouTube [8], with more signalled, including editing images found on the web [9]. So the same job title in two offices may have different capability, and the person writing the acceptable-use note cannot assume the reader can even reach the button.
The evaluation advice in the source is unusually restrained for this genre, and it is the right shape: pick a repeatable task with a clear expected outcome, compare the time to write a prompt by hand against the time to select a region and then correct the answer [10]. Note when the crop is too tight, because surrounding page content can change what the selection means [11]. That last point is a quality warning and a governance one at the same time. A user cropping for brevity is making an unsupervised judgement about which context matters, and the source is explicit that outputs still need review before they feed customer-facing, financial, legal or operational decisions [12].
For anyone running a fleet, the practical read is that the decision has already narrowed to on or off, and the interesting work is everything the toggle cannot express.
Ranked by verification strength, evidence, and original report placement.
Google has expanded Gemini in Chrome with a desktop workflow called Select from screen that lets users send a selected part of a web page directly to Gemini, drawing a box around text, an image, or mixed page content and asking Gemini to analyze or act on it in Chrome's side panel.
Google's instructions describe the process as: open the Gemini side panel in Chrome, choose Select from screen, then draw around the area to share; the chosen content is sent to Gemini as the basis for the next interaction.
Google says the selected region can contain text and/or images, so the feature is not limited to a single content type.
For Workspace domains and other managed Chrome environments, the source lists four checks before building a team process: whether Gemini in Chrome is enabled in Chrome's AI innovations settings; whether account, region, subscription or rollout conditions affect the feature; whether Workspace or other administrative controls permit Gemini in Chrome; and what internal guidance is appropriate before staff share sensitive page content with an AI assistant.
The supplied documentation confirms the Select from screen workflow but does not establish a universal price, a single worldwide availability date, or that every related Gemini in Chrome feature is included with every account tier.
Google has indicated that some Gemini in Chrome capabilities may be limited by rollout phase, region, or Google AI Pro and Ultra subscription access.
Follow any of these and your For You feed starts watching them — no settings page required.
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
Single secondary source paraphrasing vendor docs
Every claim in the cluster rests on one dev.to post that paraphrases Google instructions without linking or dating them. The workflow description is specific and internally consistent (side panel, draw region, text and/or image), and the post is candid about what it cannot establish, but there is no primary documentation, no independent confirmation, and no test or measurement of the feature's behaviour.
No adoption data supplied
The supplied material reports that the workflow exists and that availability varies by rollout phase, region, subscription, and admin configuration. It gives no user numbers, enterprise deployments, enablement rates, or dated rollout milestones, so adoption cannot be scored without inference.
Mildly overstated by framing, hedged in substance
The post's substance is unusually well hedged for a feature write-up: it disclaims accuracy, price, availability, and tier inclusion, and warns that cropped context can mislead. The overstatement is at the framing layer — a 'faster analysis' headline and an unverified expansion signal attached to a consultancy pitch, resting on zero adoption evidence and no primary documentation. Small positive gap rather than a large one.
Commercially motivated write-up of a vendor feature
The single source is a publisher post that funnels into Scalevise's paid AI consultancy with an explicit request-a-consultation call to action, and its factual base is Google's own product messaging restated without independent verification. Both the promotional intent and the vendor-origin framing are visible in the text, though the piece does also surface caveats that work against pure promotion.
Low — one commercially interested publisher
One publisher, one article, no primary documentation, no adoption data, and a disclosed-by-text commercial interest. The mechanical description of Select from screen is plausible and self-consistent enough to report, but the governance and rollout dimensions that make the story interesting are asserted at a level of generality that cannot be checked from the supplied material.
product
Google gives US students a $200 AI plan free, plus a renewal date and 5TB of switching cost3 distinct publishers
product
Relay's shutdown hands Chrome a product boss and its customers a September 14 deadline2 distinct publishers
product
Gemini reads your Workspace by default, and the off switch is in the admin console1 distinct publisher
build
Relay's last day is September 14th; its founder's next job is Chrome VP2 distinct publishers
Distinct publishers with included, body-backed reporting in this cluster.
dev.to
1 article · August 25, 2026