Overview Leaderboard Methodology
Industry Report

Sparkling Waters

Carbonated bottled waters offered plain or infused with natural flavors, positioned as low-calorie, hydrating alternatives to sodas.

Brands tracked: 35
Brands analyzed: 35
Last updated: 2026-04-23
Model: OpenAI GPT-5
Prompts: 275
Total responses: 1,400
Top Brand Overall?
Topo Chico
81/100

Highest overall AI Visibility Score in this industry.

LBA Leader?
Polar
83

Highest score on the LBA metric in this industry.

Authority Leader?
LaCroix
72

Highest score on the Authority metric in this industry.

TOM Leader?
Topo Chico
100

Highest score on the TOM metric in this industry.

Topo Chico leads the Sparkling Waters category in AI responses

Topo Chico is the most consistently-surfaced brand when users ask AI assistants about Sparkling Waters, but not to the point of monopolising the conversation. LaCroix and a handful of others appear regularly, giving users a reasonable set of alternatives to compare. The top of this leaderboard is where most of the AI-driven traffic will go.

Brand Leaderboard All 35 Sparkling Waters brands ranked

Ranked by overall AI Visibility Score (smoothed geometric mean of LBA, Authority, and TOM with an LBA-based floor on Authority and TOM, see methodology). Click any brand for the full report.

# Brand LBA Authority TOM Overall
1
Topo Chico
topochico.com
 79
67
100
81
2
LaCroix
lacroixwater.com
 76
72
79
76
3
Perrier
perrier.com
 78
61
93
76
4
Spindrift
drinkspindrift.com
 74
41
81
63
5
B
bubly
bubly.com
 73
34
55
52
6
Polar
polar.com
 83
26
60
51
7
Polar Seltzer
polarseltzer.com
 75
24
53
46
8
AHA
drinkaha.com
 74
8
57
35
9
Gerolsteiner
gerolsteiner.de
 70
7
50
32
10
Mountain Valley Sparkling Water
mountainvalleywater.com
 61
6
24
23
11
S.Pellegrino
sanpellegrino.com
 82
11
8
21
12
Waterloo
drinkwaterloo.com
 74
7
13
21
13
Liquid Death
liquiddeath.com
 71
7
12
20
14
Badoit
badoit.com
 80
8
8
19
15
SodaStream
sodastream.com
 79
8
8
19
16
Ramlösa
ramlosa.se
 72
7
7
18
17
S
Schweppes
schweppes.com
 73
7
7
18
18
DASH Water
dash-water.com
 71
7
7
17
19
Poland Spring Sparkling
polandspring.com
 69
7
7
17
20
Saratoga Spring Water
saratogaspringwater.com
 70
7
7
17
21
Vichy Catalan
vichycatalan.com
 68
7
7
17
22
Aura Bora
aurabora.com
 66
7
7
16
23
Clearly Canadian
clearlycanadian.com
 64
6
6
16
24
Ferrarelle
ferrarelle.it
 66
7
7
16
25
Hildon
hildon.com
 65
7
7
16
26
Mountain Valley Spring Water
mountainvalleyspring.com
 65
6
6
16
27
Sanzo
drinksanzo.com
 65
6
6
16
28
San Benedetto
sanbenedetto.it
 62
6
6
15
29
Crystal Geyser
crystalgeyserplease.com
 54
5
5
14
30
Hal's New York
halsnewyork.com
 54
5
5
14
31
Nixie
drinknixie.com
 48
5
5
12
32
Mineragua
mineragua.com
 33
3
3
9
33
Montellier
montellier.ca
 35
4
4
9
34
R
Rambler
drinkrambler.com
 30
3
3
8
35
V
Vintage Seltzer
vintageseltzer.com
 8
1
1
3

How is this calculated? Methodology

Every brand in this leaderboard is scored against the same set of 275 shared Sparkling Waters prompts. The same prompts, same model, same iterations. So differences in scores reflect actual differences in AI visibility, not differences in measurement.

Overall AI Visibility Score
Smoothed geometric mean of LBA, Authority and TOM. Authority and TOM are floored at LBA × 0.1 before the geometric mean (the same floor used in the per-metric cards above, so brand cards and the composite tell the same story). Formula: composite = ((LBA + 5)(Authority + 5)(TOM + 5))^(1/3) - 5. The floor keeps brands the model clearly recognises but doesn't yet recommend from collapsing to zero, while a single genuinely weak metric still pulls the composite down. Full methodology.
Shared industry prompts
For Authority and TOM, all brands in the industry are scored against the same 275 category prompts (e.g. "best SEO tools for agencies"). This makes brand-to-brand comparisons valid - everyone faces identical inputs. LBA prompts are per-brand because they ask brand-specific questions.
Latent Brand Association (LBA)
5 brand probes + 1 control prompt, each run 5 times in recall mode (no web search). LBA = quality × meta × stability × share × recognition × 100. Read the full LBA methodology →
LLM Authority
50 organic category prompts (discovery, comparison, problem and transactional intents), each run once in recall mode and once in retrieval mode. Score = frequency × log-decayed prominence × intent weight, then 50/50 averaged across the two modes. Read the full Authority methodology →
Top of Mind (TOM)
15 high-volume discovery prompts (sourced from Keywords Everywhere search-volume data), each run 5 times in pure recall mode (no web). Score = frequency × (0.5 + 0.5 × log-prominence), volume-weighted. Read the full TOM methodology →