Study: Top Google ranking gives only 49% chance of appearing in AI Overview citations

·2 views

A snapshot study conducted on July 12, 2026, analyzed 120 commercial-intent keywords across nine niches to examine how Google's AI Overviews select their cited sources. Results showed that the page ranking first organically was cited in an AI Overview only 49% of the time, suggesting traditional rankings no longer reliably predict AI visibility. However, 87% of AI Overviews did cite at least one top-three organic result, indicating high rankings remain necessary but not sufficient. YouTube and Reddit together accounted for 20% of all citations, outpacing any individual publisher or brand. The study also found only 16% overlap between AI Overview citations and those from Google's conversational AI Mode, effectively making them two separate discovery surfaces.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Developer builds browser-based CAD tool that generates 3D homes from a text prompt

A developer has launched Forge3D Spaces, a browser-based tool that converts a plain-language description into a fully furnished, walkable 3D house within seconds. The tool uses a large language model to interpret the prompt and produce a structured room specification, which a deterministic solver then converts into a geometrically consistent floor plan with no overlapping rooms or misaligned walls. The resulting model simultaneously powers 2D floor plans, 3D views, elevations, and cost estimates, so edits to one element automatically update all others. Users can further refine the design through a chat panel, issuing natural-language commands to reposition furniture, change roof styles, or add landscaping. The tool requires no installation and exports DXF files compatible with AutoCAD, aiming to replace costly drafting services for homeowners and contractors at the concept stage.

0 comments Read more at DEV Community

ProgrammingHacker News ·

Chrome 148 makes Math.tanh a tool for identifying users' operating systems

Starting with Chromium 148, the JavaScript function Math.tanh produces subtly different floating-point results depending on the underlying operating system. This variation arises from differences in how each OS implements low-level math libraries. As a result, websites and trackers can exploit this inconsistency to fingerprint users and infer which OS they are running. The issue highlights how even standard browser math operations can become unintended vectors for user identification. Security researchers have flagged this as a privacy concern for Chromium-based browser users.

0 comments Read more at Hacker News

ProgrammingDEV Community ·

CaravanAI Uses Turn-Based Voice Etiquette to Personalize Travel Planning

A developer built CaravanAI, a passion-driven AI travel companion, as part of a weekend coding challenge. The web app uses a turn-based voice interface that prevents the AI from speaking over users or listening to its own output, mimicking natural conversational etiquette. Built with Vanilla JavaScript, Vite, and the browser's Web Speech API, it requires no heavy frameworks and runs with near-zero latency. CaravanAI integrates Google's Gemini AI for itinerary generation and ElevenLabs for natural-sounding voice responses. A visual voice orb on the interface changes color in real time to indicate whether the AI is processing or speaking.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Dev Builds Blockchain Tool to Permanently Record Daily Learning Milestones

A developer has built 'Proof of Practice', a decentralized application that logs and verifies personal learning efforts using blockchain technology. Submitted as part of a weekend coding challenge, the tool lets users record practice sessions via voice or text, which are then analyzed by Google Gemini to generate coaching feedback. ElevenLabs converts that feedback into spoken audio, while progress is tracked on a streak dashboard. Verified learning milestones can be minted as permanent records onto the Solana Devnet blockchain, making private dedication publicly verifiable. The app is built with Next.js 15, TypeScript, and Tailwind CSS, and is live at proof-of-practice.vercel.app with a guest demo mode available.

0 comments Read more at DEV Community