Why Local Pack Optimization Doesn't Transfer to AI Search

Most advice about AI search visibility starts from local SEO instincts: get your GBP right, build citations, earn reviews. This is directionally correct -- those signals matter for both systems. But "directionally correct" hides a problem that shows up when businesses optimize for local pack and then check their AI search visibility and find nothing changed.

The signals are the same. The weights are not.

The Whitespark Data Separates Them for the First Time

In its 2026 Guide to Google AI Mode for Local Businesses -- a 540-query study -- Whitespark became the first source to explicitly separate Local Pack ranking factor weights from AI Search Visibility weights. Scout session 78 (filed July 15, 2026) flagged this as a significant finding for how we frame fix plans.

The two systems share the same five signal categories. But the weights diverge substantially:

| Signal | Local Pack | AI Search Visibility | |--------|-----------|---------------------| | GBP signals | 32% | 12% | | Reviews | 20% | ~16% | | On-page content | ~15% | 24% | | Directory citations | 6% | 13% | | Links | 8% | 13% |

*(Source: Whitespark 2026 Guide to Google AI Mode. Weight percentages are MEDIUM confidence; the directional finding -- GBP matters far less for AI search than for local pack, on-page content matters far more -- is corroborated across multiple independent sources.)*

GBP optimization -- completing your profile, updating hours, adding photos, earning the verified checkmark -- addresses 32% of local pack weight. It addresses 12% of AI search visibility. A business that completes its GBP has done real work. That work just matters more than two-and-a-half times as much for Maps as it does for AI search.

The signals that actually drive AI search visibility: on-page content at 24% is the top factor, followed by links and directory citations tied at 13%. Reviews land at 16%. GBP is the fifth-weighted signal in an AI search context.

A business that optimizes GBP and stops there has addressed the highest-weight local pack signal and the fifth-weight AI search signal.

The Query Type Problem Compounds This

The weight divergence is compounded by a second finding from the same Scout session: the query type businesses typically use to test their AI visibility is the query type least likely to produce AI results.

The Whitespark 540-query study found AI Overviews appear in 15% of local-intent queries -- "best plumber near me," "HVAC contractor in [city]" -- vs. 92% of informational queries and 97% of hybrid queries.

If a business owner tests "best [their service] in [their city]" and sees no AI results, that doesn't mean they have no AI visibility. That query type produces an AI result 15% of the time. They may be appearing consistently on informational queries ("how much does HVAC installation cost?" "what should I look for when hiring a plumber?") and have no idea because those aren't the queries they're checking.

This has a practical consequence for any before/after measurement: the query type needs to actually surface AI results with reasonable frequency. Local-intent queries, which feel like the obvious test for a local business, are the least likely to trigger AI Overviews.

What the Weight Data Means for Fix Sequencing

If AI search visibility is weighted toward on-page content (24%) and directory citations (13%), the fix sequence changes from what most local SEO advice prescribes.

For a business starting from zero AI citations:

**Directory presence first.** Yelp received 512,680 AI citations in Q4 2025 alone -- 3.4 times more than the next-highest directory (Foundation/AirOps study, 28.5 million AI responses, published May 2026). For local service businesses, a Yelp profile with actual reviews and a claimed BBB listing is the foundation of directory citation weight. These two cover the infrastructure the 13% citation weight is built from. And for the platforms where citation volume actually concentrates -- Google AI Mode and Perplexity together account for approximately 95% of local AI citation volume in the Foundation/AirOps data, with ChatGPT and Gemini combining for roughly 5% -- Yelp's dominance (72.5% of AI Mode citations, 62.1% of Perplexity citations) makes it the non-negotiable first move.

**On-page content second.** The 24% on-page weight for AI search rewards dedicated service pages with specific, clear content over a generic "services" catchall page. An HVAC contractor needs a page for furnace installation, a page for AC repair, a page for heat pump replacement -- each with enough specificity that an AI can parse what the business does and in what context. This content is not primarily for users clicking through a search result. It is infrastructure for AI retrieval.

**GBP third -- but still necessary.** At 12% AI search weight, GBP is not the primary lever for AI search improvement. But it is a prerequisite for Google AI Mode's local card mechanism, which pulls GBP data directly. A complete, verified GBP is required for appearing in Google's AI surfaces even though it is not the lever that drives AI Overview text citations.

The current Sourcepull fix plan architecture already sequences this correctly. The Whitespark data provides explicit external validation for that ordering.

Why This Matters If You Have Been Told to "Fix Your GBP"

A lot of AI search advice published in 2025-2026 defaults to GBP optimization as the primary recommendation. It is familiar, actionable, and GBP genuinely matters for local pack ranking. But conflating local pack optimization with AI search optimization produces the experience businesses describe: "I fixed everything they told me to and still don't appear in AI results."

They fixed the local pack levers, which are weighted differently than AI search levers. The two are not the same optimization path. The Whitespark weight data now makes that explicit.

The starting point for understanding where your specific gaps are is a baseline measurement that separates local-intent citation presence from informational query citation presence -- the query types that actually trigger AI results. Signal Check at sourcepull.ca uses informational and hybrid query framing rather than the local-intent queries that surface AI Overviews 15% of the time, which is why the results often look different from what you see when you test yourself.