This Week's Marketing Tools Dripped Out Like Toothpaste From a Tube — But Don't Rush to Buy

A while back, I scrolled past a piece of news that left me sitting there stunned.

PwC Middle East — yes, the consulting giant that spends every day telling clients "how to use AI responsibly" — put out a batch of reports on AI and electric vehicles. Sounds professional.

Then someone ran a detection tool over them —

Half the footnotes in the reports were fake.

The cited sources either didn't exist at all, or the articles in question never said what was claimed. The most absurd one: they repackaged a 2017 JPMorgan loan-review automation project as "a real-world Agentic AI success story" —

Folks, 2017. ChatGPT hadn't been born yet.

And the source they leaned on was a teenager's Medium blog with 280 followers.

"Do as I say, not as I do."

A pretty interesting business model.

The harder pill to swallow: this isn't an isolated case. GPTZero had previously exposed similar problems at EY and KPMG — both quietly retracted in the end. The very people selling themselves as AI advisors ended up with AI egg on their faces first.

I'm not bringing this up to mock PwC.

I'm bringing it up because it cuts straight to an awkward spot every marketer is staring at right now — AI tools are genuinely good, but which one should you actually be using, and what should you be guarding against?

Just How Many New Tools Popped Up This Week?

I counted. In the four weeks of July alone, the publicly announced marketing tools integrating AI numbered in the hundreds.

I dropped the list into a marketing group chat, and someone replied: "Is this a trickle, or are the floodgates open?"

Floodgates.

I picked out the ones genuinely worth talking about. Let's go category by category.

Category One: Everyone's Suddenly Scrambling for "Placement Inside AI Search"

The densest cluster of new releases this week — I'm losing count of the names — Directree's GEO Monitor, Pepper's GEO platform, Profound, SeeResponse, GenOptima, Grow Predictably, Omni Channel's OMNI 360, CrunchJunkie, Frostbite, SEO Image, Targence, Teck NOC…

What do they do?

They measure how many times your brand gets mentioned inside conversational search engines like ChatGPT, Gemini, and Perplexity, how it gets described, and whether it's being misstated.

Why the sudden burst?

Because the search entry point has changed.

It used to be that SEO meant wrestling with Google's algorithm for placement — stack the right keywords, buy enough backlinks, and you're on page one.

Now when a user asks ChatGPT "recommend a CRM for a small team," the one or two names that show up in the answer — that's the new front page.

And it's a black box. You have no idea how these large language models decide who to mention and who to skip.

So an entire new category has risen up around it. The jargon is GEO (Generative Engine Optimization), or AEO (Answer Engine Optimization).

In plain terms: the AI-era version of SEO.

Here's my call: within the next year, this category of tool will become standard kit for marketers — just like everyone scrambling to buy SEO tools a decade ago. But you have to be picky — right now everyone's methods are still crude. It's nothing more than firing prompts at the models on a timer, recording the answers, and comparing how often brand names show up. The ones that can produce genuinely actionable insight — those are rare.

Category Two: MCP — A Protocol the Marketing-Tool World Has Collectively Embraced

The second category worth talking about is called MCP.

What's MCP?

It's an open protocol that Anthropic pushed out last year, full name Model Context Protocol. You can think of it as "the USB port of the AI world."

It used to be that every AI tool wanting to grab data you'd stashed somewhere else (CRM, website backend, video library) had to write its own custom integration. Once MCP arrived, as long as a tool follows the protocol, any MCP-supporting AI client can read it directly.

This July, the ones announcing MCP servers include 6sense (B2B purchase-intent data), Guideline (global ad spend data), Webflow (website management), Wistia (video library), OneSignal (push notifications), and Omneky (ad creative).

And there's one that's quite interesting — Popl built an event-management assistant directly inside Anthropic's Claude.

What does this mean?

It means marketing tools have started to admit a fact: users aren't going to log into your backend one by one anymore. They'll sit inside AI assistants like ChatGPT and Claude, say in plain English "pull up last week's conversion data" or "send a discount coupon to this batch of customers," and let the AI call your APIs behind the scenes.

Your product is shifting from "a tool people come to use" into "a capability that gets called by AI."

That's a fundamental shift. The tools that react early will survive; the ones that don't will disappear without users even noticing they're gone.

Category Three: From "Helps You Write Copy" to "Runs Your Marketing for You"

The third category is the one that sends a slight chill down my spine.

Otter.ai put out something called Live Assist. What does it do?

While you're in a meeting, it listens in real time to your conversation with the customer, auto-matches it against a sales playbook you've uploaded beforehand, and then quietly slides a card onto your screen: Don't answer that — ask it this way.

A real-time call coach.

This isn't "AI helping you write copy" anymore. This is AI standing next to you, directing how you talk.

There are similar plays: Sprinklr letting AI agents directly handle customer-service tickets, modify user-account settings, approve refunds; Tenon embedding marketing automation inside ServiceNow; Mod Op's ORION using its own in-house models to generate ads, predict performance, and audit ad spend.

AI has gone from "content-production tool" to "marketing executor."

Sounds great, right? It does everything for you.

But think about it — that's probably what the folks at PwC thought too. AI writes the reports for us, how convenient. And out popped a teenager's blog with 280 followers as a citation source.

When an AI agent makes decisions for you, what's the actual quality of those decisions? The risk is an order of magnitude higher than when it just writes a paragraph of copy for you.

My own lean: let AI do the execution, let humans do the judgment. The other way around is where the trouble starts.

So Many New Releases — How Do You Pick?

By now you've probably sensed it: the thing most worth remembering from this week isn't any single product. It's a structural shift.

The marketing tech stack is being re-layered.

At the bottom are protocols like MCP, letting AI pull data across tools. In the middle are various AI agents, executing specific actions on your behalf. Up top are the new stuff — GEO, AEO — deciding whether you can still be seen in the AI era.

The next time another "AI-powered marketing platform" gets announced, don't rush to try it.

Ask it three questions first:

Which layer's problem does it solve?

Within that layer, who's the protocol, who's a node, who's a peripheral player?

At the judgment step, does it leave things to humans, or hand them off to AI?

Get those three things straight, and you won't get dizzy in the flood of new releases.

One last thing about PwC.

The ironic part is that, in the very same report, PwC announced a partnership with OpenAI to deploy customer-service agents for clients.

Just got slapped by AI, then turned right around and kept selling AI.

You tell me — thick-skinned, or just true believers? I can't say for sure.

But there's one thing I am sure of — this wave of AI marketing tools isn't going to stop because someone wipes out. The only thing that'll stop is the people who rushed in without thinking it through.

Picking slowly is worth more than buying in a rush.