Strategy

SEO vs LLMO

Two disciplines, one shared foundation, and a genuinely different win condition. This is the working comparison — what transfers, what's new, where old SEO habits actively cost you in AI answers, and how you measure something that has no Search Console.

What each term actually means

SEO (search engine optimization) is the practice of making a page rank highly in a list of results. The unit of competition is a page, the surface is a results page of ten links, and success is measured in position, impressions and clicks.

LLMO (large language model optimization) is the practice of making a business retrievable, accurate and quotable inside AI assistants — ChatGPT, Claude, Gemini, Perplexity, and Google's own AI Overviews. The unit of competition is a claim about you that a model is confident enough to state. Success is being named when someone asks the assistant a question your business answers.

You will see the same idea sold under other names. GEO (generative engine optimization), AEO (answer engine optimization) and AI SEO all describe substantially the same work. The vocabulary hasn't settled because the field is about three years old. Don't pay a premium for a label.

Side by side

Dimension	SEO	LLMO
The goal	Rank a link high on the results page	Be named inside the generated answer
Where it shows up	Google, Bing results pages	ChatGPT, Claude, Gemini, Perplexity, AI Overviews
Unit of competition	A page, against nine other links	A claim, against everything the model could say instead
What wins	Relevance, links, authority, page experience	Specificity, consistency across sources, structure, citability
Content that performs	Comprehensive pages targeting a keyword	Direct answers to real questions, stated as facts
Role of your own site	Primary — you rank your pages	Partial — the model also reads directories, reviews, press, profiles
How you measure it	Search Console: position, impressions, clicks	No console exists. You ask the models and record what they say
Feedback speed	Weeks to months for ranking movement	Days to weeks as models re-crawl and re-ground
Failure mode	You rank on page four and nobody sees you	You're never mentioned, and nothing tells you
Can you buy placement?	Yes — ads, clearly labeled	No. There is no slot to purchase
Who it's hardest for	New sites without authority	Businesses whose facts are vague or contradictory
Half-life of the work	Long, if content stays accurate	Shorter — models re-ground, so inconsistencies resurface

The 80% that carries straight over

Most of what a competent SEO practitioner does is exactly what LLMO needs. If someone tells you the two are unrelated disciplines, they are selling a service.

• Crawlability. A page a search engine can't read is a page a model can't read. Same constraint, same fix.
• Genuinely useful content. Thin pages never ranked and they don't get cited either.
• Clear structure. Descriptive headings, short paragraphs, one idea per section. This has always helped readers; it now determines how cleanly a model can extract a passage.
• Technical hygiene. Canonicals, sitemaps, consistent URLs, reasonable load times.
• External credibility. Being referenced by sources that carry weight in your field mattered for ranking and matters for whether a model trusts a claim about you.
• Accurate business information. Hours, address, services, pricing — the same NAP consistency work local SEO has demanded for fifteen years.

The 20% that's actually new

Where SEO habits actively hurt you

This is the part that gets left out, and it's the reason some well-optimized sites perform badly in AI answers.

• Keyword-stuffed prose. Copy written to hit a phrase repeatedly reads as low-quality to a model and produces passages not worth quoting.
• Programmatic near-duplicate pages. The city-by-city template that once picked up long-tail rankings now reads as scaled, low-value content — and gives a model no distinct fact about any location.
• Hedged, non-committal writing. "Pricing varies depending on your needs" is a sentence no model can use. A price band it can state is worth more than a page of qualifications.
• Facts locked in images and PDFs. Menus, price lists and spec sheets as graphics are invisible. This single issue accounts for an enormous share of businesses that are never recommended.
• Content gated behind a form. Whatever the lead-gen argument, an ungated competitor gets cited and you don't.
• Burying the answer to protect dwell time. Making people scroll past 800 words of preamble was always a tax on the reader. Now it also means the extractable answer never gets extracted.

Which queries this actually affects

Not all of them, and knowing which is how you avoid over-reacting. Pew Research Center, tracking the real browsing of 900 US adults across nearly 69,000 Google searches, found AI summaries appeared on 8% of one- or two-word searches but 53% of searches of ten words or more. Question-shaped searches beginning with who, what or why triggered a summary about 60% of the time.

The pattern is clear and useful: short navigational and branded queries still behave like classic search. Long, conversational, question-shaped queries are where AI answers take over. If your customers find you by typing your name, LLMO is a lower priority than someone will tell you. If they find you by describing a problem in a full sentence, it is already the main event.

What the data says about the shift

The direction is not in dispute; the magnitude gets exaggerated in both directions.

The same Pew study found that when an AI summary was present, users clicked a traditional result 8% of the time, against 15% when no summary appeared — and only 1% clicked a link inside the summary itself. Ahrefs, comparing 150,000 keywords that trigger AI Overviews against 150,000 that don't using Search Console data, measured a 58% drop in click-through rate for the top-ranking page.

Here is the honest reading of that 58%, which almost nobody includes. In the same Ahrefs analysis, comparable keywords with no AI Overview also saw position-one click-through roughly halve over the same two years. Click rates were already falling before AI Overviews existed. AI Overviews are the sharpest instance of a decade-old pattern — featured snippets, local packs, top stories — not a new phenomenon. That matters, because it tells you this is a structural drift to plan around, not an emergency to panic-spend against.

How you measure each

SEO has instrumentation. Search Console gives you impressions, average position and clicks per query, for free, with reasonable accuracy.

LLMO has no equivalent, and anyone presenting an authoritative "AI visibility score" is presenting an estimate. What actually works is unglamorous:

So which should you invest in?

The honest answer depends on where your customers actually are, not on which acronym is newer.

• SEO first if you're not yet ranking for the terms that describe what you sell, and if your buyers search in short commercial phrases. You cannot skip a foundation you never built — models draw heavily on what search already surfaces.
• LLMO first if you already rank respectably but inquiries are flattening, or if your category is one people describe in full sentences: professional services, hospitality, healthcare, anything with a considered decision behind it.
• Both, in that order, for almost everyone else — because the foundational work is shared, so the sequencing question matters less than it sounds.

What I'd avoid is paying separately for two disciplines that overlap by 80%. If a vendor quotes you an SEO retainer and an LLMO retainer, ask them which specific deliverables differ.

Is SEO dead?

No. But the version of SEO that was about manipulating rankings died over a decade ago, and the version that's about being genuinely findable and useful is now doing double duty — it's the same input that determines whether an AI can retrieve and cite you.

What is fading is the assumption that ranking equals traffic equals business. Ranking first on a query that resolves inside an AI answer produces an impression and no visit. The response isn't to abandon search. It's to stop treating position as the outcome and start treating being named as the outcome, in both places.

Where to start this week

The longer versions: what LLMO is, the playbook for getting cited by AI, and straight answers to the questions people ask consultants.