AI Mode vs Local Pack: Where Do Local Businesses Actually Appear?

When a bare keyword like "divorce lawyer" or "plumber" returned no business cards in AI Mode, adding a modifier rescued the results every time. "Near me" was the most reliable: it rescued 100% of bare-keyword failures (38/38 cases, 95% CI: 90.8%-100%).

The strongest example: bare "divorce lawyer" returned zero business cards across all 10 cities (Fisher's exact test, Bonferroni-corrected p < 0.001). Add "near me" and it returned business cards in 10/10 cities. Add "best" and it returned them in 8/10 cities. Add the city name and same result.

Criminal defense lawyer shows the reverse pattern: bare "criminal defense lawyer" returned businesses 70% of the time, but "best criminal defense lawyer" returned zero business cards across all 10 cities (BH-adjusted p = 0.037, does not survive Bonferroni correction). Adding "best" to criminal defense queries eliminated all business listings and triggered educational content instead.

When AI Mode skips business cards, it doesn't go silent. It still mentions businesses by name in the text 98% of the time, but without star ratings, review counts, or click-to-call buttons. The user sees content, not listings.

These patterns have direct implications for which content AI Mode shows. When bare keywords fail, users who naturally add "near me" or their city name will see businesses. Users who search bare keywords may get articles about the divorce process, cost breakdowns, or selection guides instead. The specific keyword-modifier combination determines whether AI Mode treats the query as informational or transactional.

This wasn't a blanket pattern for all lawyers. "Best divorce lawyer" returned businesses 80% of the time. "Best personal injury lawyer" returned them 90%. Meanwhile, "best plumber" and "best HVAC repair" returned structured business cards 100% of the time, and "best dentist" hit 90%. The intent classification varies by keyword, not just by vertical.

An important caveat: Some bare-keyword queries that returned zero businesses on day 1 were returning businesses by day 3. The structural findings (which businesses appear, review gap patterns) were consistent across all collection days. These phrasing-dependent patterns were not. Expect them to shift.

What to do: If AI Mode isn't showing businesses for one of your keywords, test it with "near me," your city name, or "best" before concluding you're locked out. The bare keyword and the modified keyword can produce completely different result types. Track both. For your Google Business Profile, make sure your primary and secondary categories match the modified versions of your target keywords, your service area is explicitly set, and your business description includes the city and service terms that trigger business cards. If AI Mode is showing educational content instead of business listings for your keywords, being the source AI Mode cites in its response may matter more than having a business card. Invest in comprehensive, authoritative content on your website that answers the questions AI Mode is surfacing.

At the median, businesses surfaced exclusively in AI Mode have 49% more reviews than those found strictly in the Local Pack (358 vs. 240, bootstrap 95% CI for the gap: +71 to +154 reviews). Businesses with fewer than 50 reviews are nearly twice as common in the Local Pack (16.8%) as in AI Mode (9.3%).

But this isn't uniform. In two verticals, the pattern reverses completely.

The consumer story: If you're a plumber, HVAC tech, or dentist, the businesses AI Mode shows have substantially more reviews than the Local Pack. AI Mode plumbers have 4x the reviews. Nearly half (47%) of Local Pack-only plumbers have under 100 reviews. Only 19% of AI Mode plumbers do.

The professional reversal: If you're a lawyer or medical specialist, AI Mode businesses have fewer reviews. These are also the verticals where AI Mode is most likely to skip business listings entirely and show educational content instead: articles about the divorce process, condition overviews, cost breakdowns.

We can't say reviews cause better AI Mode visibility. Businesses with more reviews also tend to have stronger online presence, better-optimized profiles, and more web authority. The review gap may reflect those differences rather than drive them. Mean star ratings show no meaningful difference between the two channels (4.83 vs. 4.84), so the separation is in volume, not quality.

What to do:

• Consumer services (plumbers, HVAC, dentists, chiropractors): Check your review count against the AI Mode leaders for your keywords. If there's a big gap, closing it won't guarantee AI visibility but it removes one obvious difference between you and who's showing up.
• Professional services (lawyers, medical): Reviews are less likely to be the differentiator. AI Mode appears to weight different signals for YMYL verticals: credentials, educational content, authoritative web presence. If AI Mode is showing articles instead of business cards for your keywords, focus on building comprehensive practice area pages, publishing educational guides related to your services, and earning citations from authoritative legal or medical directories.

Because AI Mode is an evolving system, a single snapshot of results is one draw from a shifting pool, not a fixed ranking.

During our study, we observed that while the overall patterns remained stable, specific business presence rates can fluctuate day to day. Our post-study validation run (47 additional API calls) confirmed the structural findings held while phrasing-dependent patterns had shifted.

What to do: Do not rely on a one-time check. Track your AI Mode visibility consistently over time. The businesses that show up across multiple checks are the ones with stable placement. The ones that appear once and disappear are sitting on the edge of the system's threshold.

AI Mode is a parallel discovery channel with its own logic. It shares about 71% of the same businesses as the Local Pack, but the 28.5% gap creates blind spots if you're only tracking traditional rankings.

What we now know that wasn't obvious before: how you phrase the query changes whether businesses appear at all, and "near me" is the most reliable trigger for business cards. The review volume gap is large in consumer verticals but reverses for professional services. And certain keyword-modifier combinations get educational content instead of business listings.

The Local Pack hasn't stopped mattering. But if you're not tracking AI Mode alongside it, you're missing businesses your competitors can see, and possibly missing that you're invisible in a channel that Google is actively expanding.

Data source: LocalSEO Data API. 1,120 total unique query executions across 7 verticals (plumbers, dentists, HVAC, chiropractors, lawyers, restaurants, medical specialists) and 13 cities, collected May 20-22, 2026.

Query design: 750 query-city pairs formed the initial dataset: 5 verticals x 3 keywords per vertical x 5 modifiers (bare, "best," "near me," "[city]," "[city state]") x 10 cities. Each query was tested in both AI Mode and Local Pack. An additional 320 queries expanded coverage to 2 new verticals (restaurants, medical specialists), 3 new cities (Los Angeles, Seattle, Boston), temporal replication, and "near me" stability testing. 50 replication re-runs checked short-term consistency. Market tier analysis combines the core 10-city dataset with extension city data; matched-pair counts reflect queries where both sources returned businesses.

Business extraction: 3,591 business mentions extracted from AI Mode responses. 659 of 750 initial queries returned structured business listings. 607 unique businesses extracted from the Local Pack.

Name matching: 8-step normalization pipeline (smart quote conversion, lowercase, legal suffix removal, separator splitting, punctuation cleanup). Two-tier matching: exact normalized (99.6% of matches), substring containment (0.3%). Automated spot-check flagged 56 potential mismatches (Jaccard >= 0.5); manual review confirmed 5 genuine missed matches, yielding a 0.36% false negative rate. Impact on overlap: +0.26 pp, negligible. Ambiguous matches conservatively treated as non-matches, meaning overlap percentages may slightly undercount true overlap.

Overlap analysis: 649 query-city pairs where both sources returned businesses. Overall LP overlap: 71.5% (bootstrap 95% CI: 69.2%-73.8%, 10,000 resamples). Rank-matched comparison (top 3 vs top 3): 48% overlap. Unique AI-only businesses: 999.

Review volume: Median review gap: AI-only 358 vs LP-only 240 (+49%). Bootstrap 95% CI for the gap: +71 to +154 reviews. Mean gap: 143% (skewed by HVAC/plumber chains). Rank-matched median gap: +88 reviews. Mean star ratings: 4.83 (AI-only) vs 4.84 (LP-only), no meaningful difference.

Modifier rescue analysis: Per-keyword presence rates calculated from 150 queries per modifier across the initial 5 verticals. Fisher's exact test with Benjamini-Hochberg correction (FDR) for all 60 bare-vs-modifier comparisons. 4 survive Bonferroni correction (all "divorce lawyer" modifier pairs, p_raw = 0.000011 to 0.000714). 5 survive BH-FDR (adds "best criminal defense lawyer" returning 0/10 business cards, BH-adjusted p = 0.037). Rescue rate confidence intervals computed via Clopper-Pearson exact method.

YMYL analysis: Bare-keyword presence rates ranged from 0% (divorce lawyer) to 100% (hvac repair, dentist). Business presence by modifier type: bare 74.7%, "best" 84.0%, "near me" 95.3%, "[city]" 92.0%, "[city] [state]" 93.3%.

1. Point-in-time snapshot. Data collected May 20-22, 2026. Structural patterns (which types of businesses appear, review gaps) held across all three days, but specific rankings can shift. Google updates AI Mode regularly.
2. Correlation, not causation. AI Mode businesses tend to have more reviews in consumer verticals, but we can't confirm whether reviews directly influence selection or reflect broader business authority.
3. API-based, not personalized. Results don't include user history or precise location signals. Real users may see different results.
4. Local Pack top 3 only. Some "AI Mode-only" businesses may appear in the expanded "More places" list below the Pack.
5. Seven verticals, 13 cities. Coverage varies by vertical and city (see n-counts in tables). Retail, real estate, and international markets were not included.
6. AI Mode adoption is unknown. Google hasn't published usage numbers, though availability continues to expand.
7. Non-independent observations. Queries in the same city share local business landscapes. Bootstrap CIs treat pairs as independent, which may slightly overstate precision.
8. Slot-count asymmetry. AI Mode shows ~5 businesses per query vs. the Local Pack's fixed 3. The rank-matched top-3 comparison (48% overlap) controls for this difference.