Sl Search Surfaces

Search surfaces: the model

TL;DR

One asset, three doors. Do not treat them as three disciplines with three budgets.

Before starting

1. Load CONTEXT.md from the repository root. It carries the vocabulary lock, the citation discipline and the freshness policy this skill assumes.
2. Establish which surface the user actually cares about. Most people say "AI SEO" and mean one specific thing: usually GEO. Ask before proceeding if it is ambiguous.
3. Check whether the question is conceptual or operational. Conceptual questions end here. Operational ones hand off in Phase 4.

Core principles

1. The surfaces are real, the disciplines mostly are not. A senior SEO doing topical depth, clean structure, schema and accurate FAQs is already executing most of GEO. Say this out loud. It is true, it builds trust, and pretending otherwise is what made this field noisy.

2. What genuinely changed is measurement, not craft. SEO measures rank. AEO measures snippet presence. GEO measures citation share inside a synthesized answer. Different instruments, different gaps, different reporting. That is why the vocabulary earns its keep, and it is the only reason.

3. Keyword stuffing inverts. The one tactic that historically worked in moderation for SEO is actively harmful in generative engines, at roughly −9% visibility against doing nothing [Aggarwal-2023]. This is the single largest divergence between the surfaces.

4. Substance beats style. In the same study the four substance tactics (adding quotations, adding statistics, citing sources, improving fluency) all cleared +28%, while tone and vocabulary tactics clustered between +6% and +18% [Aggarwal-2023]. Evidence outranks authoritative voice.

5. Classic search is still the majority surface. Around 79.8% of surveyed US consumers still preferred traditional engines for informational queries [HigherVisibility-SearchPref-2025]. GEO is the growth surface, not the current volume surface. Anyone selling GEO as a replacement for SEO is selling something.

Workflow

[Phase 1/4: Frame] Name the surfaces and kill the aliases

Three terms are used: SEO, AEO, GEO. AIO, LLMO, GAIO and SXO are aliases or product names.

Mirror the user's own term in conversation, write deliverables in the locked vocabulary, and add a one-line glossary note. Do not correct anyone publicly over an acronym.

[Phase 2/4: Locate] Work out where the user actually is

[Phase 3/4: Order] Apply the priority, and say it is a judgement

SEO first.  GEO next.  AEO last.

Because roughly half of AI citations still originate from pages ranking in the top ten, ranking is the substrate rather than a parallel track. GEO precedes AEO because the answer box is a subset of a shrinking classic SERP while citation share is growing.

Present this as a judgement call, not a finding. Someone weighting zero-click data more heavily would put AEO second and would have a case: AI Overviews have been measured reducing clicks by 58% [Ahrefs-AIO-CTR-Feb2026], and trigger on roughly a quarter of searches overall, rising to about half in healthcare [Conductor-AEO-GEO-2026].

Hard gate: if Foundations score below 50, GEO advice is withheld until that is fixed.

[Phase 4/4: Hand off] Route to the specialist skill

Best practices

• Lead with the overlap. "Most of this is SEO you already know" is the honest opening and the one that earns the rest of the conversation.
• Quantify the surface split before recommending investment, so the user is not moving budget from the majority surface to the growth surface on vibes.
• Give the free method. Manual prompt testing across four assistants costs an hour and answers most first questions.
• Date every number you present. Say "as of May 2026" out loud.

Anti-patterns

Output format

When this skill runs standalone, return:

1. Surface diagnosis, which of the three is actually at fault, with the symptom that shows it
2. The overlap statement, what the user is already doing that counts
3. Priority order for their case, with any deviation from the default explained
4. Handoff, the named next skill and why
5. Freshness note, the validation date of any figure quoted

Questions to ask when the brief is thin

1. What do you type into an AI assistant where you would expect to be named?
2. Which competitors get named instead?
3. Do you rank in the top ten for those topics in classic search today?
4. Is the site in the Bing index? (Most people have never checked, and ChatGPT search runs on it)
5. Is there a named author with credentials on your key pages?
6. Is this a YMYL topic (health, finance, legal)? If so the bar for authorship and sourcing rises.

References

Full registry: references.md.

Related skills

• sl-geo-audit, measure and score the current state
• sl-citation-content, write what LLMs lift
• sl-entity-infrastructure, make the machine able to identify you
• research/the-alphabet-problem.md, the terminology argument in full