The real shift isn’t AI content. It’s AI distribution.
Look past the noise in those headlines and a single pattern jumps out:
search, social, CRM, and even ad platforms are all quietly becoming
AI-driven recommendation systems.
Not “AI for copywriting.” Not “AI for faster dev.”
AI deciding what gets shown, to whom, and instead of what.
A few signals:
- Google AI Overviews and “AI Mode” changing how queries are answered, not just ranked.
- Sparktoro data showing AI recommendations changing with nearly every query.
- Semantic search being called “the only search that matters now.”
- Social platforms pushing social search (Bluesky SEO, Threads SEO, Discord as PR and engagement).
- AI CRM and AI-powered GA pitched as “growth engines,” not just analytics.
- OpenAI selling ads inside ChatGPT with a six-figure minimum.
The shelf your brand used to sit on (SERPs, feeds, inboxes, marketplaces) is collapsing into
one blended, AI-personalized recommendation layer.
If you’re still running a classic SEO + paid media + email + social plan without a
recommendation strategy, you’re optimizing for a world that’s already gone.
From “ranked results” to “recommended answers”
Old world: you fought for positions.
- SEO: rank in the top 3 blue links.
- Paid search: win the auction, show an ad above the fold.
- Social: post, hope the algorithm gives you reach.
- Email: send, hope for inbox placement and an open.
New world: you’re fighting for inclusion in a model’s answer.
- Google AI Overviews decide whether to cite you at all.
- ChatGPT / Perplexity decide which brands to mention by name.
- Social search surfaces profiles and posts based on meaning, not keywords.
- AI CRM decides which message, channel, and offer your customer sees next.
The question has changed from “How do I rank?” to:
“Why would this model choose me as the safest, most useful recommendation right now?”
The operators’ problem: invisible loss and noisy data
The Ahrefs work on tracking AI Overviews and “traffic Google won’t show you” is the canary in the mine.
You’re losing:
- Clicks to AI answers that satisfy the query without a visit.
- Attribution clarity as journeys route through AI surfaces you can’t tag.
- Control as models rephrase, compress, or omit your brand entirely.
At the same time, tools like Google Analytics are being sold as “growth engines,”
AI CRMs promise “revenue-driving use cases,” and WordPress ships AI agents to “speed development.”
Everyone’s giving you more knobs to turn, while the shelf your product sits on is being rebuilt.
That’s why a lot of teams feel the same thing:
you’re optimizing the funnel while the store is moving.
Think in “recommendation surfaces,” not channels
Stop thinking “SEO vs social vs email vs ads.”
Start thinking in surfaces where recommendations happen:
- Answer surfaces: Google AI Overviews, ChatGPT, Perplexity, Gemini, Copilot.
- Discovery surfaces: TikTok, Reels, YouTube, Threads, Bluesky, Discord servers.
- Owned decision surfaces: your site search, product recommendation blocks, CRM/ESP, in-app prompts.
- Paid recommendation surfaces: search ads, social ads, retail media, ChatGPT ads, third-party endorsements in search ads.
Each surface has a model (or several) making a call:
“What’s the lowest-risk, highest-utility thing to show this person right now?”
Your job is no longer just to “create content” or “bid correctly.”
Your job is to make your brand the default low-risk recommendation on the surfaces that matter.
The 5 levers of a modern recommendation strategy
Here’s how to operationalize this shift without burning your team down.
1. Model-friendly relevance: semantic, not just keyword
Ahrefs is right: semantic search is the game now.
Models don’t care if you repeat the exact keyword; they care if you fully resolve the intent.
Practical moves:
-
Cluster by intent, not by phrase.
Build one strong, canonical asset per intent (problem, solution, comparison, implementation),
then support it with related content. Kill or consolidate cannibalizing pages. -
Write for questions, not tags.
Use headings and body copy that mirror real questions: “How much does X cost?” “Is X safe for Y?” -
Cover the full decision path.
Don’t just chase “best [category]”; cover “alternatives,” “vs,” “for [segment],” “for [use case].”
Models like sources that let them answer follow-up questions without leaving.
2. Trust signals for humans and machines
Social Media Examiner’s “Does AI trust you?” is the right question.
Models are trained to avoid embarrassment. If recommending you feels risky, you’re out.
Trust, in this context, is a mix of:
- Authority: clear authorship, expertise, and real-world proof.
- Consistency: no wild swings in claims, pricing, or positioning.
- Alignment: your content matches what users actually do next (low pogo-sticking, decent dwell, conversions).
Practical moves:
-
Add structured credibility.
Author bios with credentials, case studies with numbers, citations to third-party data.
Use schema where it’s still useful (FAQ, how-to, product, org). -
Clean your claims.
Remove outdated stats, vague superlatives, and conflicting numbers across pages.
Models are increasingly good at spotting contradictions. -
Invest in third-party proof.
Google testing third-party endorsements in search ads is a tell.
Reviews, analyst notes, partner logos, and social proof all make you a safer pick.
3. Format for extraction, not just reading
AI Overviews, answer boxes, and chat models need to extract your value.
If your content is a wall of prose, you’re harder to quote.
Practical moves:
-
Use “quotable blocks.”
Clear definitions, numbered steps, pros/cons lists, comparison tables.
Think: “If a model grabbed this block, would it stand on its own?” -
State the answer early.
First 2-3 sentences should directly answer the query, then expand.
That’s what answer engines want to lift. -
Design for skimming.
Subheads that match search and chat queries, short paragraphs, bullets.
This helps both humans and models understand topical coverage quickly.
4. Feedback loops: stop flying blind in AI surfaces
You can’t fully measure AI Overviews or ChatGPT mentions, but you can get directional signal.
Practical moves:
-
Set up an “AI visibility” review.
For your top 50-100 queries, manually check:
AI Overviews, People Also Ask, chatbots (where possible), social search.
Track: “Mentioned? Cited? Ignored?” -
Correlate with click loss.
When you see stable rankings but traffic drops, flag those terms as “AI-risked.”
Prioritize improving content depth, trust signals, and format there. -
Instrument downstream.
If top-of-funnel visibility is opaque, make sure mid- and bottom-funnel analytics are tight:
better UTM discipline, clean CRM data, and conversion tracking that actually works
(recall the “73% of ecommerce emails are broken” stat – assume similar rot elsewhere).
5. Own your recommendation engines
While everyone obsesses over Google and OpenAI, you quietly control high-value recommendation systems already:
your site, app, and CRM.
Tools are moving fast here: AI CRM use cases, AI GA, AI site search, product recommenders.
Used well, these are not just personalization toys; they’re your hedge against external algorithm volatility.
Practical moves:
-
Fix the basics before adding AI.
If your current recommendations are generic (“popular products”), don’t expect AI to save you.
Start with rule-based segments and clear hypotheses. -
Design a “house model” roadmap.
Where will you let AI decide next best action?
Examples: onsite search results, product sort order, email send-time and content, in-app prompts. -
Guardrail with business logic.
Don’t let models recommend things you can’t fulfill or support.
Hard-code inventory, margin, and compliance rules around AI systems.
Media buying in a recommendation-first world
Paid media is quietly shifting from “renting impressions” to “buying into recommendation engines.”
A few implications for media buyers and growth leads:
-
Creative becomes the model’s training data.
Platforms like Meta talk about “turning dials” – they’re optimizing off your creative and conversion data.
Run more creative variety with clear positioning, not just more variants of the same ad. -
Context and endorsement will matter more.
Third-party endorsements in search ads, creator integrations, and UGC that models can see
all contribute to your “safe to recommend” profile. -
Stop worshipping last-click ROAS.
As more discovery happens in opaque AI surfaces, your cleanest metrics will be branded search,
direct, and downstream conversion efficiency. Treat them as signals of recommendation success. -
Test the new shelves early.
ChatGPT ads with a $200k minimum aren’t for everyone, but they’re a preview.
Look for smaller, similar surfaces (AI assistants, vertical AI tools, retail media search)
where you can buy early signal at sane budgets.
What CMOs should actually change in the next 12 months
You don’t need a 40-slide “AI strategy” deck. You need a few hard decisions.
-
Reframe your KPIs around recommendation, not just reach.
Add metrics like “share of answer” (how often we appear in answer-like contexts),
“branded search growth,” and “assisted conversions from discovery channels.” -
Fund one cross-functional “recommendation squad.”
Small team spanning SEO, content, media, CRM, and analytics.
Mandate: improve the brand’s odds of being recommended on 3-5 priority surfaces. -
Kill low-signal content and campaigns.
If a page, sequence, or campaign doesn’t clearly support an intent cluster or recommendation surface,
stop feeding it budget and headcount. You’re training noise into the system. -
Invest in trust infrastructure.
Reviews operations, case study production, data hygiene, brand safety policies.
These are boring, but they make you a low-risk choice for humans and models. -
Decide your AI risk posture.
Where will you aggressively adopt AI (owned recommendations, analytics, ops)?
Where will you be conservative (brand voice, sensitive segments)?
Make this explicit so teams aren’t paralyzed or reckless.
The operators who win the next few years won’t be the ones who generate the most AI content.
They’ll be the ones who understand a simple, uncomfortable fact:
you’re no longer just marketing to people – you’re marketing to the systems that decide what people see.