AI Search Killed Your Press Releases (Proof Inside)

AI search engines have already decided your press releases don’t exist — and four million citations prove it. Meanwhile, OpenAI is building a self-serve ad platform inside ChatGPT, and a LinkedIn executive just shipped iOS apps without writing a single line of code. The distribution layer, the creative layer, and the advertising layer of marketing are all being restructured simultaneously, and the window for first-mover advantage in each is measured in months, not years.

Your Press Releases Are Invisible to AI Search — The Data Is In

A dataset of four million AI-generated citations confirms what many suspected but nobody had measured: syndicated press releases register at near-zero citation rates in AI search answers, while editorial content from owned newsrooms performs significantly better. The entire PR wire model — Cision, PR Newswire, Business Wire — was engineered for a world where broad syndication equaled broad reach, and AI search engines have invalidated that assumption almost overnight. If your audience discovers information through ChatGPT Search, Perplexity, or Gemini, and those engines don’t cite your announcements, your announcements are effectively not happening.

Redirect PR budget toward building owned newsroom content and original editorial voices this quarter — that is the content type AI citation engines are actively rewarding right now, and the advantage window before everyone else catches on is short.

Read the full story →
Join the discussion →

ChatGPT Is Building a Full Ads Manager — Marketers Should Pay Attention Now

Code strings inside the ChatGPT Android beta — “ad target,” “search ad,” and “search ads carousel” — describe the scaffolding of a self-serve advertising platform that puts OpenAI on a direct collision course with Google Search. This isn’t a banner ad experiment; it’s the infrastructure for a new paid media channel concentrated in free-tier users, who represent ChatGPT’s highest volume audience. The critical open question isn’t whether ChatGPT can serve ads — it clearly can — but whether it can build the conversion attribution infrastructure that performance marketers require before shifting serious budget.

Start internal conversations about ChatGPT budget allocation scenarios now, because the first movers on a new ad platform historically capture disproportionate ROI before competition normalizes pricing and auction dynamics.

Read the full story →
Try it yourself →
Join the discussion →

LinkedIn’s Editor-in-Chief Shipped iOS Apps With Zero Coding Experience

Daniel Roth, LinkedIn’s editor-in-chief, used a dual Claude agent architecture — one agent builds, one reviews for security and quality — to ship fully functional iOS apps without any prior coding experience, a story published through Lenny Rachitsky’s network. The builder-plus-reviewer pattern is production-ready and directly replicable for marketing automation, content pipelines, and campaign QA workflows right now. The critical buried detail: Claude expressed false confidence about a deprecated library, and only a manual Reddit check caught the error — meaning silent AI failure, not loud AI failure, is the failure mode practitioners most need to guard against.

Experiment this week with a two-agent Claude setup where one agent produces output and a second critiques it — and build mandatory human verification checkpoints into any workflow that touches customer data or live campaign infrastructure.

Read the full story →
Try it yourself →
Join the discussion →

Claude Can Now Read Your Figma Files Directly — A Hidden Workflow Accelerator

The Claude-to-Figma “Code to Canvas” integration, launched February 17, 2026, lets Claude Code read Figma files directly — extracting design tokens, components, layouts, and dev mode snippets in a single session. For marketing teams running parallel design and content workflows, this means AI-generated copy and code can reference live brand design systems simultaneously, collapsing the design-to-development handoff from days to hours. This integration has received almost no marketing-practitioner coverage despite being a free capability upgrade for anyone already inside both ecosystems.

If your team uses Figma for campaign assets and is already experimenting with Claude, test the Figma MCP integration immediately — extracting design tokens directly into an AI prompt is a non-obvious workflow acceleration hiding in plain sight.

Read the full story →
Try it yourself →
Join the discussion →

AI Search Engines Are Describing Your Brand — And You Probably Don’t Know What They’re Saying

A Czech SEO specialist’s documented workflow using Semrush’s AI Sentiment Insights shows that monitoring how AI search engines describe your brand — and correcting inaccuracies — is now a distinct and necessary marketing discipline, not an optional audit. AI search engines synthesize brand descriptions at scale from sources you don’t control, and those descriptions are what millions of users see as authoritative answers; getting that description wrong in an AI answer is worse than a negative review because it is invisible to the brand and trusted by the user. Brands with complex, frequently updated products — enterprise software, SaaS platforms, D2C retailers with deep catalogs — are especially vulnerable.

Run your brand name through Perplexity, ChatGPT Search, and Gemini this week, document exactly how each describes your company and products, and treat any discrepancies as urgent correction priorities rather than minor inconsistencies.

Read the full story →

Scaling AI Content Volume Without Governance Is the Same Mistake, Again

Pedro Dias at Search Engine Journal makes the case that the current AI content scaling wave is simply the latest iteration of the “publish more pages” playbook — a cycle that ended in Google penalties in 2012, 2016, and 2022 without exception. The velocity of AI-enabled content production makes the eventual correction potentially more severe this time, not less, because the scale of disappointment being accumulated is orders of magnitude larger than in previous cycles. The counterargument — that AI can genuinely improve quality-per-unit if the editorial governance layer is rigorous — is valid, but only if teams actually build that governance layer before scaling, which most don’t.

Before launching any AI content scaling program, audit the quality threshold you’re scaling to — volume without genuine differentiation is a liability, not a strategy, and the historical pattern suggests most teams will ignore this warning until the penalty arrives.

Read the full story →
Join the discussion →

Google Is Testing Store-Level Shopping Ads — The Playbook Is About to Change

Google is testing “Sponsored Shops” blocks in Shopping results that promote entire store identities rather than individual products — a structural shift that would move retail advertising competition from “which product wins the auction” to “which store earns the placement.” If this rolls out broadly, brands with strong catalog depth gain compounding advantages over single-product advertisers, and brand authority signals become more valuable than individual product bid efficiency. The deeper philosophical shift is significant: Google would be moving from selling performance to selling presence.

Watch this test closely and start building the internal business case for brand-level Shopping investment now, because if Sponsored Shops launches broadly, the entire Google Shopping optimization playbook will need to be rebuilt from the store-identity level up.

Read the full story →
Try it yourself →
Join the discussion →

NVIDIA’s $1 Trillion Backlog Is a Signal About Content Strategy, Not Just Chips

At GTC 2026, Jensen Huang revealed a one-trillion dollar NVIDIA sales backlog projected for 2027, alongside new AI infrastructure products including the OpenClaw architecture and Vera CPU — numbers that describe enterprise AI infrastructure commitment at a scale that is not a temporary experiment. For marketing technologists, the non-obvious read is that a compute constraint at this scale could create meaningful access and pricing disparities between enterprise and SMB users of premium AI models well into 2027. The connection to the AI citation data earlier in this report is not coincidental: NVIDIA’s backlog is the infrastructure bet that AI search becomes the primary information interface, and the citation dataset is early proof that bet is being won.

Factor AI infrastructure permanence into your three-year technology strategy — the organizations treating AI tooling as a provisional experiment are operating on a fundamentally false premise about the longevity and depth of this transition.

Read the full story →
Join the discussion →

Federal AI Preemption Kills State Regulation — Compliance Strategy Needs to Shift

A Florida AI regulation bill backed by Governor DeSantis failed to advance after President Trump signaled federal opposition to state-level AI oversight, creating a clear preemption signal that the patchwork state-regulation nightmare scenario for enterprise AI programs is now significantly less likely. For marketing teams that had been building compliance workflows around potential state AI frameworks, the risk calculus has shifted materially — but reduced regulatory friction also means competitors will move faster and the brand differentiation play shifts from “we comply” to “we choose to be trustworthy,” which is a harder story to tell and a more durable one to build. EU AI Act implementation timelines now deserve more compliance attention than US state-level tracking.

Redirect any state AI compliance workflow energy toward tracking federal policy signals and EU AI Act implementation timelines, and begin building the internal narrative around proactive trustworthiness rather than reactive compliance.

Read the full story →

OpenAI Codex Security Closes the Last Gap in AI-Only Code Workflows

OpenAI’s Codex Security tool uses AI-driven constraint reasoning and validation instead of traditional static analysis to identify real vulnerabilities with significantly fewer false positives — a meaningful departure from legacy security tooling that previously required specialized expertise to act on findings. For marketing technologists shipping AI-generated code in workflows, customer data pipelines, or integrations, this closes the loop on a workflow that now runs from design extraction (Figma MCP) through code generation (Claude Code) to security review (Codex Security) without requiring traditional technical specialists at any stage. The caveat applies here too: AI-driven security tools are themselves AI systems with potential blind spots, and the risk of over-trusting AI security review at this stage is real.

Evaluate Codex Security as a review layer for any AI-generated code your marketing team is shipping — particularly relevant if you are experimenting with the dual Claude agent workflow described earlier in this report.

Read the full story →

Kagi Built a “LinkedIn Speak” Translator — And It Reveals a Real Opportunity

Kagi Translate now supports “LinkedIn Speak” as an explicit output language, earning 185 upvotes on Hacker News — a community signal that LinkedIn’s corporate communication style is culturally distinct, widely recognized, and formulaic enough to be satirized with a dedicated tool. For marketing practitioners building LinkedIn-based creator strategies, the fact that the platform’s communication norms are legible enough to parody means differentiation there is both more achievable and more valuable than most practitioners realize. The counterargument — that LinkedIn’s algorithm may actually reward the formulaic patterns because engagement behavior has been conditioned by years of that content — is worth holding alongside the opportunity.

Use the Kagi LinkedIn Speak translation tool to stress-test your LinkedIn drafts against the clichés it’s designed to produce — if your content sounds like the output, rewrite it before publishing.

Read the full story →
Try it yourself →
Join the discussion →

Watch the Full Video Breakdown

I cover all of these developments in my daily YouTube video, including live demos of the tools mentioned above.
Watch today’s full breakdown on YouTube →