From Semantics to AI: The Same Debate, Repackaged

If there is one thing that has not changed since the days of the I-Search Mailing List, it is SEOs' ability to disagree. Back then, it was debate. Today, it is often dismissal. Right now, the industry is splitting again, this time over AI.

The Recurring Narrative

We have seen this pattern before. For nearly three decades, the phrase "SEO is dead" has resurfaced repeatedly. The first notable pronouncement came as early as November 1997, when Richard Hoy declared on the Online Advertising Discussion List that search engines were a "dead-end technology" and that fretting over rankings was a waste of time. He advised clients to create good meta tags, submit their sites, and then forget about it.

Yet, search never died. It evolved.

By 2005, entrepreneur Jeremy "ShoeMoney" Schoemaker similarly declared the end of SEO, arguing that search engines were improving so rapidly that short-term top rankings would inevitably be corrected. In 2009, Robert Scoble proclaimed that SEO was getting dramatically less important, suggesting that SEM should be renamed to "OM" for Online Marketing. Danny Sullivan, responding on Search Engine Land, noted the cyclical nature of these claims, stating, "I've covered the space going on 14 years now. I've heard the SEO is dead spiel over and over and over again."

What we are seeing now is simply a new version of the same cycle:

• GEO vs SEO
• AI vs Search
• Traditional vs Generative

One camp claims SEO is obsolete. The other claims nothing fundamental has changed. Neither position holds up on its own. Both are, in different ways, incomplete.

The False Binary

Historically, major algorithm updates have consistently triggered panic. The Florida update in 2003 and the Cassandra update in 2004 marked the end of the "Wild West" era of keyword stuffing and link spam. The Panda (2011) and Penguin (2012) updates led to headlines declaring that Google was making the SEO industry obsolete. When Hummingbird arrived in 2013, shifting focus to user intent rather than just keywords, many declared keyword-centric SEO finished.

Access the SEO Founders’ User Guide — a curated knowledge base shaped by experienced SEO practitioners. Move beyond theory and into practical, proven strategy.

What changes is how information is retrieved, interpreted, and presented. Not whether information matters. As noted during the Panda and Penguin updates, "SEO may not be dying, it is certainly changing and changing fast." The methods deployed to trick algorithms become irrelevant, but legitimate practises around content quality and structure remain critical.

Early SEO and the Emergence of Semantics

In the early days of search, practitioners were already writing for something that had not yet been formalised. We did not call it "semantics" at the time, but that is what it was. Writing naturally with synonyms, structured explanations, descriptive language, and contextual clarity was essential. Even when search engines like Google struggled to interpret intent, good writing still followed logical principles. Why? Because it was obvious where things were heading.

Then vs Now

Then (weak intent understanding)

Consider the query: "best running shoes for flat feet". Content often looked repetitive: "Best running shoes for flat feet are important if you have flat feet. These running shoes for flat feet provide support." Search engines struggled with synonyms like "trainers" or "support shoes", context such as budget or injury, and intent layers spanning informational, commercial, or comparative needs. The result was repetitive content, exact-match dominance, and thin SERPs with little differentiation.

Now (semantic + intent-aware)

The same query today is interpreted differently. Modern content might say: "If you have flat feet, stability or motion-control trainers can help reduce overpronation. Look for structured arch support, firm midsoles, and a secure heel counter." What has changed is that synonyms are understood, concepts are mapped (flat feet to overpronation), and pages can satisfy multiple intents simultaneously.

The Logical Progression

We did not guess this direction. We reasoned it. It was logical that systems would evolve to better understand meaning, relationships, and context. And they did. The same logic applies to where we are now with AI.

Where We Are Today

It is important to ground this in reality. AI systems are not finished products. They still struggle with attribution and source clarity, consistent entity resolution, weighting true expertise versus repetition, freshness and evolving authority, and blending information without explicit provenance. In many cases, they still confuse visibility with credibility.

But these are not permanent limitations. They are transitional ones.

The Shift We Are Already Seeing

People often say, "AI traffic is only a small percentage compared to traditional search." That may be true today. But the direction of travel matters more than the current distribution. Adoption curves tend to follow the same pattern: early slow adoption, rapid acceleration, and eventual rebalancing of channels.

We have seen this with mobile, voice, and other interface shifts. There is no reason to assume AI will behave differently. Especially when major platforms, including Google, are actively integrating AI-driven experiences into their core products.

What This Means Across Different Industries

The shift to AI-driven inclusion affects different sectors in distinct ways. Understanding your specific context is critical.

B2B SaaS: Entity-Based Authority Over Brand Authority

For SaaS companies, the shift is particularly significant. Historically, a strong brand (backed by venture funding, media coverage, and ranking visibility) could dominate search results. In an AI-driven environment, that changes.

Example: A mid-market CRM competitor might have weaker brand visibility than Salesforce, but if their documentation, case studies, and thought leadership are more consistently cited, more specific, and more reusable across AI systems, they'll be included in more AI-generated answers. A prospect asking "How do we migrate from Salesforce to a more affordable alternative?" might see the mid-market player cited alongside Salesforce, not below it.

What changes: You're no longer competing just for ranking position. You're competing for inclusion in synthesised answers. That requires different signals: entity consistency, citation velocity, and information reusability.

E-commerce: Reviews Become Synthesis Material

E-commerce sites have always relied on reviews for trust and ranking. But AI systems are changing how reviews function.

Example: A consumer asks ChatGPT, "What's the best lightweight hiking boot for narrow feet?" The AI might synthesise information from 5-10 sources: manufacturer specs, Reddit discussions, professional reviews, and user testimonials. A brand that has consistent, detailed reviews across multiple platforms (their own site, Amazon, outdoor forums, YouTube) will be cited more frequently in AI-generated answers than a brand with reviews only on their own site.

What changes: Reviews aren't just for conversion anymore. They're synthesis material for AI systems. Brands that distribute their review content strategically across platforms where AI systems can access it will see more AI-driven traffic.

Professional Services: Thought Leadership Becomes Verifiable

For consultants, agencies, and professional service firms, thought leadership has always mattered. But AI systems evaluate it differently.

Example: A management consultant publishes an article on "Why most digital transformations fail." In a traditional search environment, ranking depends on backlinks, domain authority, and keyword optimisation. In an AI environment, the system asks: "Is this information cited elsewhere? Do other experts reference it? Can I verify the claims independently?"

A consultant whose ideas are cited by other practitioners, referenced in case studies, and corroborated across sources will be included in AI answers about digital transformation. One whose thought leadership lives only on their own site will be marginalised.

What changes: Thought leadership must be distributed and corroborated, not siloed. Your ideas need to appear in other people's content, not just your own.

The Real Debate Being Missed

Much of the industry is still arguing around the wrong questions: Does EEAT exist? Does schema matter? Is GEO legitimate? These are surface-level disagreements.

The more meaningful question is: What signals does a system require to confidently reuse and synthesise information? Because that is where search and AI are converging. Not on rankings. But on inclusion.

From Ranking to Inclusion

The shift is subtle but significant. Historically, success meant ranking positions. Moving forward, success means being selected as part of an answer.

This introduces a different optimisation problem. It is not just about visibility, but recognisability, consistency, and trust at an entity level.

What "Authority" Becomes in This Context

Authority is no longer just a page-level concept. It becomes entity-based, time-sensitive, and distributed across signals. Systems will increasingly evaluate who is behind the information, how consistently they appear across sources, and whether their contributions are corroborated elsewhere. In other words, authority becomes something that must be inferred, not declared.

How to Adapt: 5 Immediate Actions

Understanding the pattern is one thing. Acting on it is another. Here are five specific steps you can take right now to position yourself for this shift.

1. Audit Your Entity Consistency

Start by mapping where your organisation, brand, or name appears across the web. AI systems use entity resolution to understand "who is this?" Your goal is consistency.

Action: Search for your name, brand, or company across:

• Your own properties (website, blog, social)
• Third-party platforms (LinkedIn, industry directories, review sites)
• News and media mentions
• Academic or professional databases

What to look for: Inconsistent descriptions, outdated bios, conflicting information, or missing context. If your LinkedIn bio says one thing and your website says another, AI systems notice.

2. Test Your Content in AI Systems

Don't wait for traffic data. Test your content directly in ChatGPT, Perplexity, Claude, and Google's AI Overviews.

Action: Ask these systems questions your content answers. Examples:

• "What's the best approach to [your topic]?"
• "Who are the experts in [your field]?"
• "What's the latest thinking on [your subject]?"

What to observe: Are you cited? If not, why? Is your content too promotional? Too vague? Not specific enough? This direct feedback is invaluable.

3. Map Your Information for Reusability

AI systems need to be able to extract, synthesise, and cite your information. That requires structure and clarity.

Action: Review your key pieces of content and ask:

• Can this information be extracted without context and still make sense?
• Are claims specific enough to be verifiable?
• Is there a clear source attribution?
• Can this be combined with other sources without confusion?

What to improve: Add structured data (schema markup), clarify your unique perspective, and remove ambiguity.

4. Distribute Your Ideas Across Platforms

Your best ideas shouldn't live only on your website. They need to appear in places where AI systems can find them and where other practitioners cite them.

Action:

• Publish key insights on LinkedIn, Medium, or industry platforms
• Contribute to industry publications or guest posts
• Participate in discussions where your expertise is relevant
• Encourage others to cite and reference your work

What to measure: Citation velocity—how quickly your ideas are referenced elsewhere. This is now a key signal.

5. Build Citation Relationships

Authority in an AI-driven environment is about being cited, not just ranking.

Action:

• Identify practitioners and publications in your space
• Share their work and reference it in your own
• Create content that naturally invites citation
• Build relationships with people who might cite you

What to avoid: Manufactured citations or artificial link schemes. AI systems are sophisticated enough to detect these.

Early Adopters and the Advantage Curve

Early adopters have always benefited the most in shifts like this. Not because they had certainty, but because they acted on direction. Right now, dismissing AI optimisation entirely often reflects confirmation bias, comfort with existing models, or protection of current positioning.

But ignoring a directional shift does not stop it from happening. It only delays adaptation.

Those who start these five actions now will have a 6-12 month advantage before this becomes table stakes. By 2027, every serious content strategy will include AI-driven inclusion optimisation. By then, the early movers will have already established entity authority and citation patterns that are difficult to replicate.

The Uncomfortable Reality

Most content, regardless of quality in a traditional sense, fails a more modern test: Does it add anything meaningfully new? If the answer is no, systems have less reason to prioritise it. This applies across both search and AI-driven retrieval.

This is not a criticism of existing content. It is a recognition that the bar has shifted. Content that was sufficient for ranking may not be sufficient for inclusion in AI-generated answers.

Where This Leads

We are moving toward systems that retrieve information, evaluate sources, synthesise responses, and select what to include based on trust, consistency, and usefulness. In that environment, visibility alone is insufficient. What matters is whether your information is distinct, credible, and reusable.

Ready to Go Deeper?

This article establishes the pattern and the direction. The next article, "Programmatic SEO in the Age of AI: From Scale to Signal," dives into the tactical implementation: how to structure content, optimise for AI inclusion, and build the systems that make this sustainable at scale.

The five actions above are your starting point. The next article is your roadmap.

Final Thought

We have been here before. Before semantics had a name. Before intent was understood. Before systems could interpret meaning at scale. Back then, those who adapted early benefited the most. The same pattern is repeating now. Just with new terminology.

The question is not whether AI changes search. It already is. The question is: What is the minimum set of signals required for a system to confidently reuse your information? Because that is the direction everything is moving toward.

And the answer to that question is what separates early adopters from those playing catch-up.