AI-referred traffic is now converting at measurably higher rates than traditional web traffic, and the data is solid enough to act on. Add to that ChatGPT Ads getting product feed infrastructure as they continue to expand their platform, Facebook rewriting how discovery works inside the app, and six ad platforms connecting to AI agents through Model Context Protocols (MCPs). Here are the five most impactful stories from the month of June for online advertisers.
1. AI-Referred Traffic Is Converting at Higher Rates
Source: Reuters | June 15, 2026
The performance gap between AI-referred traffic and traditional web traffic is now measurable, and the numbers are significant. A Reuters analysis of retail and e-commerce conversion data found that shoppers arriving via AI sources (ChatGPT, Perplexity, Google’s AI Mode, and similar platforms) convert at rates approximately 42% higher than average web visitors and spend 48% more time on-site. Adobe Digital Insights’ Q2 2026 report states that AI-referred site traffic has grown 393% year-over-year across its retail client network. First Page Sage’s research on ChatGPT conversion rates by industry adds further detail: B2B software categories are seeing the highest lift, followed by financial services and home improvement. The explanation behind the data makes intuitive sense: users ask an AI assistant a specific question, receive a recommendation, and then click through with their intent already refined. They’re not browsing. They’re completing.
| JumpFly Takeaway A 42% higher conversion rate is not a rounding error. The first practical move is getting visibility into LLM traffic to your site. If you’re not already segmenting AI-referred traffic separately from your organic and paid channels in Google Analytics, you should start now (organic might be automatic, but paid ads need to be sourced). You can’t optimize what you can’t see. Beyond measurement, this reinforces why SEO still matters, and matters more than ever. Strong SEO is the foundation for strong GEO. Google itself even stated this at Google Marketing Live last month during their keynote. The AI systems recommending your brand are reading the same signals as search engines: structured product data, clear service descriptions, real reviews, and authoritative content. SEO is what truly drives GEO. The brands that have done the SEO work are the ones winning the AI organic results right now. And with the wild conversion rate numbers, you definitely need to be there. Get your SEO right. We can help. |
2. Facebook Adds an AI Mode To Bring Conversational Search Into the App
Source: Meta Newsroom | June 15, 2026
Meta announced a significant expansion of AI capabilities within the Facebook app on June 15, with the most notable addition being an “AI Mode,” a dedicated conversational AI experience built directly into the platform’s search and discovery layer. Rather than returning traditional links, Facebook’s AI Mode surfaces answers drawn from the platform’s own content: public Groups, Reels, and community posts, synthesized and delivered in a conversational format powered by Meta AI. The announcement also included tools to help users manage events, coordinate group activities, and get local recommendations, all through a conversational interface that keeps users inside the app rather than routing them to external sites. Meta is positioning this as a direct response to the way users are already using AI assistants as their default discovery tool and as a way to deepen engagement with Facebook’s existing content ecosystem. This is a meaningful shift in how the platform routes discovery, and it has direct implications for how brand content surfaces.
| JumpFly Takeaway The part worth paying attention to isn’t the feature itself; it’s what it signals about where discovery is heading on every platform. Facebook is training its users to expect conversational answers instead of scroll-and-find. I, for one, am a doom scroller, so my need for AI Mode in Facebook means nothing as of today (but I’m sure that will change in the future). For advertisers, the near-term impact is that your brand’s presence in public Facebook Groups and your organic content on Reels isn’t just a community play anymore; it’s input data for the AI answering people’s questions inside the app. That also makes it yet another location where your brand content needs to be monitored for consistency. What’s showing up in Groups and Reels can surface in AI Mode responses, so if that content doesn’t reflect your brand accurately, it’s a problem. Whether Facebook eventually layers paid inventory into this experience is a question of when, not if; Meta always monetizes. This only means more potential for us advertisers in the foreseeable future. What form that takes, and what kind of results it delivers, we’ll have to wait and see. The practical move now is treating your Facebook organic content strategy as AI-readiness work and making sure what’s already out there actually reflects your brand well. |
3. OpenAI Adds E-Commerce Infrastructure With Product Feeds to Its ChatGPT Ad Platform
Source: OpenAI Ads Developer Portal | June 2026
OpenAI expanded its advertising platform with product feed capabilities in June 2026. The company announced this update directly to advertisers via email and shared details on the OpenAI Ads developer portal. The update adds three meaningful capabilities: advertisers can now upload a product catalog directly into Ads Manager, build campaigns from that feed rather than creating individual ads manually, and benefit from improved conversion matching through “user objects,” a new attribution mechanism that connects purchase signals back to ad exposures more accurately than click-based tracking alone. The product feed infrastructure is a significant step toward making ChatGPT Ads a viable e-commerce performance channel rather than a brand awareness play. For context, ChatGPT processes an estimated 50 million shopping queries daily, and the platform’s existing ad inventory had been limited primarily to contextual text placements; the feed capability moves it meaningfully closer to the performance channel model advertisers are already running on Google and Meta.
| JumpFly Takeaway We covered ChatGPT Ads in our previous blog, AI in Online Advertising: 5 Key Trends from May 2026, and watching OpenAI build out the platform has been instructive. No conversion tracking, then conversion tracking, and soon the ability to optimize for conversions, and now product feed support. OpenAI has been systematically adding the infrastructure that performance advertisers need, and doing it fast. Product feeds are significant because they move ChatGPT Ads from a text placement to something that can actually compete as an e-commerce performance channel. The 50 million daily shopping queries flowing through ChatGPT now have a direct path to your catalog. Making it easier for users to purchase is the right call in my opinion. I love e-commerce. I love optimizing Shopping and Performance Max campaigns, and this is absolutely something I will be testing in the very near future. |
4. Gemini Omni in the Real World: My Tests Over the Last Month
Source: Google Blog | May 19, 2026 (real-world testing: June 2026)
Google introduced Gemini Omni at Google I/O 2026, a video generation and editing model that takes input from any combination of text, images, audio, and existing video and then responds through natural conversational language. The defining capability is multi-turn editing: each prompt builds on the last, characters stay consistent across changes, and the scene retains context throughout the session. This allows users to change environments, add or remove elements, shift visual style, or alter what’s happening in a clip without returning to a production set. The model also includes improved physics reasoning, handling realistic movement, gravity, and fluid dynamics. Gemini Omni Flash is rolling out now to Google AI Plus, Pro, and Ultra subscribers through the Gemini app and Google Flow, with free access on YouTube Shorts and YouTube Create. All outputs carry Google’s SynthID watermark.
| JumpFly Takeaway I tested Gemini Omni inside Google Flow using the Gemini agent, and the results are worth sharing in detail. For b-roll content and bringing still product images to life, the output was genuinely good: quality was strong, style was controllable, and the speed was real. It did take some tweaking and multiple iterations, but the result was usable. The gap showed up when we introduced people. Characters looked great on the first pass, but minor inconsistencies appeared, like character consistency within the same scene it generated. There was also an instance where users were holding red wine, and then suddenly it became white wine. When we prompted the model to fix a specific inconsistency, it addressed some issues while introducing others. The fix wasn’t reliable. The trust-but-verify rule applies here: review everything before it goes live, especially anything with people in it. For b-roll, product animation, and environmental scenes, it’s worth getting into now. Just don’t skip the review step. |
5. Six Ad Platforms Now Have Official MCP Servers
Source: Pinterest Newsroom + Microsoft Advertising Blog | June 17, 2026
Pinterest and Microsoft Advertising each launched official Model Context Protocol (MCP) servers on June 17, 2026, bringing the total number of major ad platforms with first-party MCP servers to six, alongside Google Ads, Amazon Ads, Yahoo DSP, Meta, and TikTok. MCP, originally developed by Anthropic and now governed by the Linux Foundation, is a standardized connection layer that lets any compatible LLM/AI agent connect directly to a platform and read live campaign data without custom API development for each model. Both Pinterest and Microsoft launched as read-only, meaning AI agents can pull performance data, account health signals, and optimization insights, but cannot change bids, budgets, or campaign settings. Google, Amazon, and Yahoo took the same read-only-first approach; Meta and TikTok are the exceptions with full read/write capability.
| JumpFly Takeaway So how can an MCP help? Instead of manually logging into six dashboards to pull reports, an MCP-enabled agent can read across Pinterest, Microsoft, Google, Amazon, and others in a single session and surface what needs attention. It provides quick analysis and recommendations through whatever LLM you’re working with, which is very useful. But the LLM doesn’t know your client history, your account nuances, or whether a flagged budget pacing issue is actually worth acting on. That’s where the marketer has to stay in the loop. Trust but verify. The speed of AI analysis is real, and it does allow for quicker review and faster identification of opportunities, but so is the expertise required to evaluate what the AI is actually telling you. At JumpFly, we have our own proprietary tool to help with analysis and recommendations, but we do not blindly accept what the LLM tells us. It is very good and surely helps all of us out on a day-to-day basis. Knowing what to do with an AI recommendation is still a marketing skill learned through experience. |
Looking Ahead
The common thread in June isn’t the technology; it’s the judgment required to use it well. Knowing when AI video output is good enough versus when it needs another pass. Knowing whether an MCP recommendation is worth acting on or just plausible-sounding. Knowing that SEO isn’t less important in the age of AI, it’s actually more important. The tools are genuinely useful, and the speed gains are real. None of that removes the need for someone who knows what they’re looking at. That’s been true through every platform shift for the past 20 years, and it’s still true now.
Questions about how these AI trends affect your campaigns? Let’s talk.
