Benoit Petry

Growth Strategist | Google Ads, LinkedIn & Content Strategy | $11M+ in ad spend | Scaling businesses with strategic, data-driven marketing

[Benoit Petry @petry_benoit]

Information/content is a data layer for agents.

Spending time, being present and visible on platforms where humans are - is becoming of capital importance.

[Matthew Prince 🌥 @eastdakota]

Welp, that happened faster than I predicted. Thought it would be end of 2027, then early 2027, but agentic traffic growing so fast that bots have now passed human traffic online for the first time in the Internet's history. https://t.co/2zX5bHdhsa

[Benoit Petry @petry_benoit]

@peeplaja Does it mean that the majority (53%) of mid-market and enterprise B2B companies did NOT cut marketing roles in the last 12 months (no headcount reduction, etc.)?

[Benoit Petry @petry_benoit]

No. I use Reddit on daily basis and it’s not a typical informational.

It’s more Word of Mouth at scale.

It’s easy to spot self-promotion/affiliate content and redditors are quite harsh to what they feel is low value or disguised advertising.

As long as Reddit stays true to it core value (i.e independent communities) it has a bright future.

[Benoit Petry @petry_benoit]

@sytaylor They will come up with efficient and affordable models.

[Benoit Petry @petry_benoit]

@TimOverbooked Most executives are obsessed about sales metrics; they don’t think about what are the steps that lead to sales.

[Benoit Petry @petry_benoit]

@dramaricic is that local SEO, national or international?

[Benoit Petry @petry_benoit]

@chris_nectiv Great example of the marketing team not talking to the product team.

[Benoit Petry @petry_benoit]

@YoungbloodJoe They have no choice but to dominate this new technology - and making sure they can insert ads in it.

[Benoit Petry @petry_benoit]

I tested Schema on 55 pages… and AI started citing them more.

It’s not a magic fix or silver bullet, but I saw a clear positive impact.

Sure, it could be correlation rather than causation, but there’s literally zero downside to adding proper schema.

Read @thinking_slow's post about @ahrefs study to make your own opinion: https://t.co/bPtrVTkVZG

[Cyrus | SEO AI @CyrusShepard]

https://t.co/CSGrp8icxQ

[Benoit Petry @petry_benoit]

@patrickstox When Google began emphasizing E-E-A-T, they knew exactly what they were doing.

[Benoit Petry @petry_benoit]

Absolutely. Schema markup is not a silver bullet for getting recommended by AI.

That said, providing structured data in a way that AI systems can easily access and understand can only be beneficial.

The main limitation I noticed in the Ahrefs study is that it didn’t specify the depth or quality of the schema implementations they tested.

It’s like claiming that working out provides no health benefits after monitoring 2,000 people, without ever examining the intensity, consistency, or quality of their training programs and diet.

[Benoit Petry @petry_benoit]

The biggest practical difference was this:
no-agent = broader strategy
agent = tighter operating system

[Benoit Petry @petry_benoit]

I ran the same strategic prompt through Grok twice: once with a dedicated agent, once without.

I expected the difference to be obvious.

It wasn't. But one version was clearly more useful and the reason surprised me.

The prompt: Build a practical ABM strategy for a specialized B2B marketing agency targeting larger mid-market clients with limited resources.

Not a generic marketing plan. A real business prompt where quality depends on constraints: small team, limited budget, need for pipeline, realistic execution.

To avoid judging on vibes, I built a scoring rubric across 8 criteria:

▸ Prompt fit and constraint handling: 15 pts
▸ ICP and account selection quality: 15 pts
▸ Messaging and positioning depth: 15 pts
▸ Content and channel strategy: 15 pts
▸ Execution feasibility: 15 pts
▸ Measurement rigor: 15 pts
▸ Clarity and readability: 5 pts
▸ Evidence discipline / overclaiming: 5 pts

Final scores:
Grok with agent: 88/100
Grok without agent: 78/100

The difference wasn't that the agent version sounded smarter.

It was more constrained.

The no-agent version gave a solid ABM Lite strategy.

But it recommended a heavier scope: 50 target accounts initially, 20–30 in the pilot, a 90-day timeline.

Not bad. But for a resource-constrained agency, it's already a lot.

The agent version was tighter: 15–25 accounts max, 5 Tier 1 accounts, one 8-week pilot, 10–15 hours/week for the marketer, scale only after 1–2 wins.

Something a small team could actually execute.

The sharpest difference was in account scoring.

The no-agent version had a solid model. But the agent version added "Accessibility" as a criterion, whether key decision-makers (CMO, Head of Growth, VP Marketing) are actually reachable.

A company can be a perfect ICP fit with strong intent signals. But if you can't reach the right person, it's not a realistic ABM target.

That one detail made the whole answer more operational.

The positioning was sharper too.

Instead of generic agency framing, the agent version landed on:

"Too big for freelancers, too small for big agencies."

One line. Tells you who it's for, who it's not, and why it exists. That's a market wedge, not a tagline.

The no-agent version wasn't weak though.

It explained the 1:Few ABM Lite model clearly and included a useful 70/30 content principle, 70% reusable across clusters, 30% custom. Solid thinking the agent version missed entirely.

So this isn't agent = good, no agent = bad.

The real difference: the agent version was better at turning strategy into an operating system.

Smaller list. Clearer pilot. Better workload assumptions. More realistic milestones.

It felt less like a good answer and more like something you could use on Monday morning.

Here's where I'll be honest.

Both versions made unsupported performance claims like win rates, ROI targets, pipeline benchmarks stated as fact without evidence.

The agent version was actually more confident about this, not less.

Constraint handling doesn't fix overconfidence.

Strong AI outputs still need a human filter, especially in marketing, where models produce benchmarks that sound authoritative and aren't backed by anything.

My takeaway:
The value of agents may not be better writing.
It may be better constraint handling. Less generic advice. More realistic sequencing. More operational detail.

For business use cases, that matters more than elegance.

[Benoit Petry @petry_benoit]

Here’s the scoring rubric I used.

[Benoit Petry @petry_benoit]

@downdetector Same here. Grok is unresponsive.

[Benoit Petry @petry_benoit]

@JulianGoldieSEO Yes

[Benoit Petry @petry_benoit]

@neoviky @georgekh @doganuraldesign @nikitabier oh... I did upgrade to premium plus, but I had also access to X Pro while premium few months ago...

[Benoit Petry @petry_benoit]

@georgekh @doganuraldesign @nikitabier What do you mean? I still have access to X Pro.

[Benoit Petry @petry_benoit]

@chris_nectiv This is huge - screenshot! I wonder why and when agent would decide to make a screenshot. It’s extremely important to have a site that is correctly linked internally.

[Benoit Petry @petry_benoit]

@YoungbloodJoe It would have been interesting to check how detailed were the schema - whether they were all about the same level of depth.

These pages could have used very light schema. Maybe that’s for another test.