The Feed Report: Issue #5
Google handed us keyword data for our Instagram posts and we're dancing
Platform ‘big ones’ this week happened while most of us were either chest-deep in a dirty lake or drenching our hot-dogs in condiments.
On July 7, Google Beyoncé-style-surprise-dropped Search Console into social analytics. No press, no keynote, just a blog post from a content writer and suddenly clouds opened, sun beams shined and you can suddenly see Google search queries pulling people to your Instagram grid. Hallelujah.
Reddit also started cracking the whip on AI by using it’s own AI (okay, lol) to hunt down citation-seeding tactics the industry has been billing clients for.
Aaaaand we finally got the real numbers on where AI sits in the purchase funnel, which turns out to be not even close to checkout. Cool.
Burn Notice is up top like always. Skim it, know it happened, get on with your day. The three stories that change how you actually work this week are down in The Safe.
Let’s ride.
Burn Notice
Free for everyone. Ten things that happened. Most need nothing from you beyond knowing they happened.
Meta added updated disclosure tags for AI-generated ads
Ads built or heavily edited with Meta’s gen-AI tools (Background Generation, Image Generation, Add Animation) now get an AI info label automatically, tucked under “About this ad” in the three-dot menu. Meta also reads C2PA metadata to catch third-party tools like Photoshop and DALL-E. The labeling happens without you, so there’s nothing to do. It’s worth knowing because undisclosed AI content is already one of the top three reasons ads get rejected this year.
X built a native iOS video editor to push original content over stolen reposts
I don’t always include X updates because most of the time they’re a waste of time, but this one caught my eye.
Caption overlays in multiple languages, Green Screen, segmented recording. Product head Nikita Bier pitched it as a way to reward original creators, and pointed out that video is now about half of all impressions on X. iOS only for now. If X isn’t in your mix, file it and move on. Worth a raised eyebrow: there’s still no built-in way for creators to report stolen work, which Meta and YouTube both have.
They’re starting to build a foundation for more original content on X, which is something I’ll keep an eye on, but for now it feels like they’re sick of seeing IG/TikTok reposts dominate views on-feed.
Facebook is testing view counts on posts
Only the creator sees them, which lines Facebook up with Instagram and Threads, where views has already become the metric that matters. Meta confirmed the test and gave no timeline. Nothing to do here. It’s just more evidence that Meta wants “views” to be the number everyone judges reach by, everywhere.
Instagram’s Edits app got auto-translated bilingual captions
Note: This is outside of this week’s news window, but I forgot to include this one last week, but it’s an important one to be aware of if you’re working internationally.
Fifteen languages, plus overlay support in templates, a clip-lock feature, and summer sound effects. It connects to Instagram’s push to surface translated content in Explore. If you’re cutting Reels in Edits for a multilingual or international audience, this is a genuine reach lever. Everyone else, good to know it exists.
The Supreme Court is letting Texas enforce its app-store age-verification law, for now
In a one-sentence order, SCOTUS declined to pause SB 2420 while the case plays out. The law makes app stores verify ages and tie under-18 accounts to a parent who has to approve downloads and purchases. The 5th Circuit hears the merits in August. No ad-account action, but if you market apps or aim at younger audiences, this is the age-gating friction that keeps spreading state by state.
Google started a small test of ads in AI Mode, beginning with healthcare.
Note: This one broke in early June, but some fresh coverage landed this week that it’s still in test mode.
Limited US and English test. Eligible campaign types: Performance Max, AI Max with search-terms matching, Shopping, and broad match. The first version leaves out creatives that use pinning or text disclaimers. Right now it only matters if you’re a US healthcare advertiser. For the rest of us it’s the tell for how AI Mode monetization reaches every regulated vertical next.
Google Ads updated its Terms of Service, effective July 1.
The new terms give Google explicit permission to use automated and AI features to format, pick, or generate your targets, ads, and destinations for you, while you keep full responsibility for whatever comes out. The expanded language also covers how your inputs get used, including what you type into conversational tools and any URLs you let Google crawl. Nothing to do. It’s just worth knowing where the responsibility lands as the automation keeps widening.
Google is testing black sitelinks in a bigger font on desktop
With a divider, in Search ads. It’s a cosmetic UI test. Know it happened, that’s the whole story.
Google launched PMax Channel Diagnostics
More visibility into how Performance Max is doing at the channel level. Incremental, but a real transparency win if you live in PMax.
Cloudflare’s new crawler rules could catch Google’s crawlers too.
An infrastructure story I’d not normally cover, but something to keep an eye on with downstream reach implications as AI-crawling and content-access rules tighten up. No action, but keep half an eye on it if you spend any time thinking about how your content gets accessed and surfaced.
The Safe
Paid-exclusive. Three stories worth your actual attention this week, translated to the account level with what to do about them.
Search Console is also a social analytics tool now
On July 7, Google launched a new Search Console property type called ‘Platform Properties’. You connect an Instagram, TikTok, X, or YouTube account and see how that content does in Google Search and Discover. No website required, no domain to verify. You authorize the account and it shows up in the property selector right next to your sites.
For years, social handed you in-app analytics. Views, likes, saves, the whole vanity buffet. What it has never shown you was how people found you through Google. That query data is the single most useful organic signal we’ve ever had for websites, and now you get it for your Instagram grid. Real terms, real clicks, real impressions, filterable by individual post and by query, and you can export the whole thing.
Last year, it was big news when Instagram posts started appearing in Google search, and it now is the complete link of marketing harmony to see how everything connects without owning the domain. Pure heaven.
Three reports come with it. Performance is the one that matters for the data-divas: total clicks and impressions, sortable by post and query, exportable. Insights covers traffic trends and the headlines for leadership and non-social team members, your top posts, and how people discover the account. Achievements tracks click-threshold milestones over a rolling 28 days, and you can safely ignore it.
Before you get excited, read the fine print, because two quirks will bite you in a client report if you walk in cold.
A click still counts when a video plays inside Google’s own viewer instead of opening the app. Instagram Stories log as impressions when they surface and as clicks when someone taps them. So “clicks” here does not mean “people who landed in the app,” and you want to know that before a client asks.
The second one is sneakier, and it’s the one that makes you look careless if you miss it. The Insights summary card counts clicks across web, image, video, and news search. The detail lists underneath it only cover web search. Your totals will not reconcile. That’s how it’s built, not a bug and not something you fumbled in the export. Drop a note in your report template now so future-you doesn’t lose an afternoon chasing a gap that was never real.
What Google is actually doing here is turning Search Console into one command center for visibility. AI Overviews and AI Mode reporting are already folded in. Social and video are the next layer. As discovery keeps sliding off homepages and into feeds, Google wants to own the measurement for all of it, and this pairs with the public Search profiles they rolled out in June. Platform properties are the private, analytics-only twin of those.
The read I keep coming back to: Google is making itself the referee of cross-surface discovery data at the exact moment the rest of us are scrambling to figure out where our organic reach even comes from.
What to do:
Check whether the property type has reached your account yet. The rollout is gradual, over “the coming weeks.”
Once it lands, connect every platform each brand is active on.
Audit which platform earns the most Google Search impressions. The answer is regularly not the one the client would guess.
Mine the query data for caption and content ideas, and tune bios, titles, and hashtags around terms already surfacing your posts.
Add the summary-vs-detail caveat to your report template today.
When the client asks “does this replace our native analytics?” It doesn’t. Native shows in-app behavior. This shows how Google Search sent people to the content, which native tools can’t see at all. Set the expectation that the data starts thin and fills in as the rollout reaches you.
What we don’t know yet: full country availability, whether more platforms get added, whether access shifts by account size or category, and how far back any historical backfill goes. Google hasn’t said.
Your GEO play on Reddit just got riskier, and Reddit’s own AI is doing the hunting
Reddit announced on July 6 that its improved automated systems are now catching roughly 25,000 spammy posts and comments a day, which cut user exposure to that content by about 20% year over year. The target has a name: generative engine optimization. Marketing content planted in Reddit threads so that ChatGPT and Gemini, both of which license Reddit’s content, will quote it back as if it were a real person’s opinion.
Reddit is using LLMs to catch the subtle, coordinated patterns older systems used to miss, and it’s screening signals right at account creation. Human moderators still did the bulk of the work, driving more than half of removals through the back half of 2025, but the automated net just got a lot finer.
Then there’s the Cornell finding, which explains the whole mess. Researchers found that a snippet as short as 13 words in a Reddit comment could sway an AI-generated answer. In one test, a single planted comment pushed a fake dating app to the top of the recommendations. Thirteen words. That’s why brands started doing this, and it’s also why it’s nearly impossible to fully police.
Reddit is stuck in a trap it built itself. The licensing deals with OpenAI and Alphabet turned it into one of the most-cited sources in AI answers, and that citation value made it the most attractive target on the internet for stealth marketing. So Reddit has to protect the integrity of the corpus it’s selling, or the deals lose their worth.
So, let’s call it what it is: defensive infrastructure with a community-health story bolted on the front. Meanwhile the money keeps flowing toward GEO shops, with Profound crossing a $1B valuation after a February round. The tactic is professionalizing at the exact moment the platform starts punishing it.
This is organic and earned-media strategy. If you or a client have been seeding Reddit for AI citations, the risk math changed this week, and suspensions are ramping.
What survives is real participation: becoming an actual member of a subreddit, building a niche community, earning citations the slow way. One marketer in the coverage has clients become moderators and build their own subreddits, taking the takedown risk as the price of doing it honestly. Play the long game or don’t play.
What to do:
Audit whether any current client’s “AI visibility” strategy leans on Reddit seeding, and flag the changed risk this week.
Move toward real community participation instead of planted content.
If you’re working inside niche subreddits, watch account health closely.
Reframe the client conversation around durable presence, not exploits.
For the client chasing “get us mentioned in ChatGPT”: they need a reset, and this is the week to give it. The line I’d use: seed-and-pray now risks bans and wasted spend, and the durable play is real presence. It’s also a clean way to separate your practice from the agencies still selling the shortcut.
What we don’t know: Reddit’s false-positive rate, how much manipulated content still slips through and gets cited before takedown, and whether OpenAI or Google re-filter anything Reddit removes after the fact. We have Reddit’s side of the numbers, and only Reddit’s side.
AI owns the top of your funnel now. Search still owns checkout.
Similarweb dug into thousands of shopping journeys across three industries and came back with the cleanest behavioral picture we’ve had of where AI actually sits in the buying process. People lean on AI chatbots at the start, for product discovery, weighing options, getting their bearings. Then they go back to traditional search to compare prices and make the final call.
The pull of that first AI recommendation is not small. Brands mentioned by tools like ChatGPT saw 2.5x more visits. And the people who arrived from AI viewed nearly twice as many pages and spent double the time on-site compared to search visitors. That engagement gap is the part I’d underline. AI isn’t just sending traffic, it’s sending people who stick around and dig in.
But it stops at the deal. Shoppers still don’t trust AI at the moment of purchase, so they route back to search to check prices and buy. Which quietly retires the “AI is replacing search” narrative and hands you something more useful: two surfaces doing two different jobs.
This is the first solid data putting a number on AI’s spot in the funnel instead of the usual hand-waving, and it maps cleanly onto the Reddit story above and the healthcare-ads test down in Burn Notice. The whole ecosystem is repricing around AI as a discovery surface, and now we have proof of exactly where that surface ends. The deal-finding gap is the interesting part, because it says AI hasn’t earned trust at the transactional moment, and that tells you where your budget and your optimization should still point.
What to do:
Map your current content to funnel stage, and be honest about the split between what’s built for discovery and what’s built for conversion.
Get your product data clean and consistent enough that AI tools surface it accurately. That’s the new version of optimizing for discovery.
Keep price and comparison content anchored where search shoppers convert.
Watch the deal-finding gap. It may close as AI shopping tools mature, and if it does, the whole map moves.
For the client: this is useful for resetting anyone stuck at either extreme, the “AI is killing search” panic and the “AI is a fad” shrug. The data backs a both-and funnel strategy. Skip the conversation entirely if they’re not e-commerce or retail.
What we don’t know: the coverage never named the three industries studied, so category behavior may vary. And whether the deal-finding trust gap holds or just melts away as these tools grow up.
That’s the week.
Three stories that all point the same direction, which I didn’t plan and only noticed writing it: Google, Reddit, and a pile of shopping data all repricing themselves around AI as the front door and search as the register.
If the Search Console social property has already shown up in your account, go connect everything and hit reply to tell me which platform is winning your Google impressions. I want to know if anyones actually lines up with what the client swore it would be. I’ve seen some gnarly keywords driving to creator’s X and Instagram pages.
See you next Thursday. Xoxo,
Mckenna