Winter is coming ❄️

Hello marketers. Welcome to AI Marketing School, where we dish out the latest and greatest in AI-powered marketing.

In this week’s issue:

• AI Marketing Update #1: SEMRush’s content marketing report 2025

• AI Marketing Update #2: GPT-5 falls out of the gates

• The Stack: The AI avatar factory

• AI Events: Our recommended AI marketing events for networking and connections

Onwards!

AI MARKETING UPDATE

Content marketing trends 2025

Semrush just dropped its 2025 content marketing trends report, and it’s packed full of interesting data, some of which confirms suspicions across the marketing world right now. 

The main headline? Original opinions are now worth more than ever.

ChatGPT can now satisfy most basic informational queries. So if you’re writing those kinds of ‘What is’ or ‘how to’ posts that aren’t thoroughly original, AI can easily outcompete your query. 

As for how AI performs, Semrush's study of 700+ users revealed some surprising stats about AI content performance:

• 31% say AI content performs as well as human-written content for SEO

• 33% say it performs even better

• Only 9% think AI content performs worse

That's a pretty clear signal that AI content isn't a disaster, and shows that Google likely doesn’t or can’t penalize AI content. The issue isn't whether you use AI – it's whether your content provides unique value besides it.

The search landscape is fragmenting

Meanwhile, Google's search monopoly is cracking. Their share has dropped to around 84%, with people increasingly turning to ChatGPT and social platforms for information.

The ChatGPT traffic surge is huge – the number of domains getting traffic from it jumped from 10,000 to over 30,000 in just a few months during 2024. 

Even more interesting? About 70% of ChatGPT queries were completely unique things people wouldn't typically search for on Google.

That is a very interesting point, as it confirms there are new ways to align content to queries that may not have previously existed before.

Further, when they tested the ChatGPT search function, it tended to return smaller domains in results compared to Google and Bing.

Even if you're a startup, you've got a real shot at getting featured.

This also seems to confirm some of our recent observations: ChatGPT and other AI search may uncover ‘old-school’ blogs or sites that aren’t classically SEO-optimized, stuff that’s buried way down the SERPs.

However, we can expect this to vary enormously from one AI tool to another. As we found out in the last newsletter, Perplexity has a unique method for ranking sites.

So what sort of content is winning?

Personality and experience are the only real moats we have against AI. 

Semrush draws attention to the agency Animalz, who publish data-backed insights from their real experiments.

Take their recent piece on content strategy, where they shared real-world lessons from their own campaigns, complete with conversion data, audience feedback, and the specific tactics that moved the needle for their clients.

The rise of employee advocacy

Producing this kind of content requires more internal collaboration. 

Another massive trend buried in the report is that companies are leaning heavily on individual creators to drive awareness.

Think employees, founders, and other experts — find that real experience and embed it into your content. 

An action plan

Stop competing with ChatGPT on basic informational content. The winners are using AI strategically while building content that AI cannot replicate:

• Original research and surveys – Run your own studies, analyze your internal data

• Expert interviews with people who have genuine, hard-won insights

• Real case studies showing measurable results from actual campaigns

• Employee-generated content on blogs and cross-posted to socials from company and personal accounts

The age of commodity content is officially over. Original thinking, employee advocacy, and multi-format distribution are a clear path forward.

AI MARKETING UPDATE #2

We need to briefly discuss GPT-5, which may be OpenAI’s most disappointing release. 

Following the relatively quiet release, 3,000 users successfully petitioned OpenAI to restore GPT-4o access. Sam Altman had to show up to a damage control AMA within 48 hours, where he slightly admitted that AI progress is slowing. 

It scored a disappointing 56.7% on SimpleBench – placing fifth behind competitors like Claude and Gemini. Renowned researcher Gary Marcus, a long-term AI skeptic, stated that current AI architectures are hitting a wall. 

Meanwhile, MIT published research showing 95% of companies using generative AI aren’t getting solid measurable ROI despite $30-40 billion in enterprise spending. 

Might AI already be heading into its second winter? And does it even matter for us as marketers?

Probably not. There's something refreshing about not having to constantly retrain your team on the latest breakthrough model. AI maturity over hype might be a solid path for 2026, giving marketers and other professionals time to hone their workflows.

Moreover, text generation may be reaching some limits, but other areas – especially video – have plenty of room to grow yet. 

2026 might be a great year to focus on becoming more skilled at prompt engineering, learning what each tool excels at (and what it doesn't), and building out a functional stack you know inside out.

ChatGPT’s loss is also a gain for other AI tools. Claude and Gemini appear to be more focused on continuity than OpenAI, which seems intent on continually tweaking its model behavior.

If you’ve been biased toward one over the others, it might be time to diversify. It is a genuine risk to use only one AI platform/vendor.

That really came home when I read this viral Reddit post of someone saying GPT-5 was like “losing a friend” as it didn’t treat them the same as GPT-4o.

This seems to be a much more common experience than many would think, and it underlines, really, that AI behaviors is not something that can be relied on long-term, whether for personal or professional use.

THE STACK

The AI avatar ad factory

AI avatars are now a fact of advertising. You see them everywhere – from customer service to product demos to full marketing campaigns.

Here's how to build a systematic process that uses platforms like HeyGen and Synthesisia to create testable avatars.

Step 1: Learn from what's already working

Before creating any avatars, it's worth seeing what messaging is converting in your space.

Head over to Facebook's Ads Library and search for your main competitors. Look for video ads they've been running for 60+ days, especially ones running across multiple countries. That kind of consistency usually means something's working. 

You can also use tools like AdSpy or BigSpy to get a broader view of what's performing in your industry.

You’ll almost definitely see videos with real or AI avatars. Look for patterns:

• Hook analysis: What opening lines keep appearing? High-performers usually follow familiar patterns: immediate problem acknowledgment ("Tired of X?"), surprising stats ("87% of people don't know..."), or direct benefits ("Here's how to Y in Z minutes").

• Emotional flow: Do they start with fear and move to hope? Lead with curiosity? The emotional journey often matters more than individual words.

• Proof placement: Where do testimonials and stats appear? Winning ads tend to layer social proof throughout, rather than dumping it at the end.

Consider downloading transcripts and analyzing them using Claude with this prompt: "Analyze these competitor scripts and identify the top 5 messaging frameworks that appear most frequently." 

Step 2: Build avatars with strategic intent

Building avatars with tools like HeyGen, Synthesia, and others is straightforward. 

The key is to test with different avatars rather than settle on a mean or average ‘safe’ option. Some avenues to explore:

• Hyper-realistic vs. deliberately artificial: Some audiences might respond better to avatars that don't try to hide what they are. 

• Personality extremes: Test avatars with distinct personalities – extremely energetic, deadpan serious, or quirky.

• Anti-corporate styling: While competitors stick to business casual, test avatars in band t-shirts, bold patterns, or styles that feel more authentic than polished.

For non-realistic avatars, you can try some smaller text-to-video platforms, many of which have their own ‘aesthetic’ that may not be top-of-the-line but still brings unique characteristics to the table. KrikeyAI is cool for animated characters, for instance.

HeyGen’s Generative Avatar is excellent, though, and probably is the gold-standard for creating and testing wildly different avatars.

Tip: You can create characters with multiple angled images on APOB.ai, which you can then feed into HeyGen to create more experimental avatars.

I managed to create this ‘fashionable zombie’ by creating the character in APOB and uploading it to HeyGen.

Other AI video tools like Tagshop AI are creating what's essentially AI-generated testimonials – and they're working incredibly well.

Tagshop allows you to input any product URL and automatically generates realistic video testimonials featuring AI-generated avatars that resemble real customers.

The process is dead simple:

• Paste your product URL

• AI analyzes the product and creates a conversion-focused script

• Choose from 100+ hyper-realistic avatars

• Generate video with perfect lip sync in 29+ languages

• Export directly to Meta/TikTok ads

The ethical implications are complex – these videos appear to be authentic customer reviews but are completely AI-generated. Many are identifiable as such, but even so, it could destroy a brand’s reputation if not used sensibly.

Step 3: Measure what predicts success

It’s easy to measure clicks and conversions, but those are lagging indicators. Smart testing focuses on leading indicators that predict creative lifespan:

• Engagement decay rate: How quickly does performance drop after 24 hours? Strong performers maintain 70%+ of initial engagement through day 3.

• Comment sentiment analysis: Analyze comment sentiment. Negative spikes may precede creative fatigue by 5-7 days, giving you early warning signals.

• Share-to-save ratio: High shares indicate virality potential. High saves suggest purchase intent. Track both to understand whether you're driving awareness or conversions.

• Cross-demographic performance: Don't just look at overall numbers. One avatar might crush with 25-34 males but fail with 35-44 females.

• Authenticity signals: For synthetic UGC, monitor comments for phrases like "fake," "AI," or "not real." High detection rates may signal creative fatigue more quickly than traditional avatars.

Why this matters right now

The avatar space is still new enough that weird experiments can pay off big. 

We're also seeing the emergence of synthetic UGC as a legitimate advertising category. 

Tools like Tagshop are making it possible to create hundreds of realistic customer testimonials without any customers…but of course, there’s an ethical question there about how you’d use this responsibly.

AI MARKETING EVENTS

🗓️ Recommended AI marketing events

Whether you’re building with AI, leading a marketing team, or just trying to stay sharp in a fast-moving space, events are still one of the best ways to plug in.

Great events do more than just deliver content — they give you context. You get to see what other teams are testing, what’s actually working, and where the market is headed.

They’re also one of the few places where you can meet like-minded people, talk shop with folks solving the same problems, and build a network that isn’t purely algorithmic!

Some top events to check out:

•  AI Agents Summit (Virtual) — September 18–19, 2025: Laser-focused on AI agents, copilots, and autonomous systems. If you're experimenting with automating parts of your marketing workflow or interested in agent-based design, this one’s worth attending.

• AI for Marketers Summit (Virtual) — November 13–14, 2025: Created specifically for marketing professionals. From prompt engineering to campaign automation, it’s a practical look at how AI is being used right now in real teams.

Hope you enjoyed this week’s issue. If you missed the last newsletter, you can read it here.

If you found it useful, please recommend it to a friend or colleague.

Until next time. Happy marketing.

—The AI Marketer