Tower 28 AI Market Strategy Report - Clean Makeup Brands

Answer Capsule

Tower 28 holds third position in clean makeup AI recommendations with 36.3% valid recommendation coverage in September 2026, down 0.8 points from July 2026. The brand is surfaced in half of all qualified observations but converts that presence into top-three placement at a rate below the category leaders. Its clearest strength is a strong presence rate of 50.4% with no negative framing across tracked platforms. Its clearest weakness is a low rank-one rate of 2.5%, indicating the brand is frequently mentioned but rarely chosen first. The clearest opportunity is converting its high reference rate into stronger top-of-shortlist placement through targeted recommendation-stage content.

Who This Report Is For

This report is for brand, digital, and growth leaders at Tower 28 and competitive strategy teams tracking AI-driven discovery in the clean makeup category.

Report Card

Executive Summary

Tower 28 holds a stable third-place position in clean makeup AI recommendations with 36.3% valid recommendation coverage in September 2026, down 0.8 points from 37.1% in July 2026. The brand peaked at 39.2% in August 2026, meaning its September figure represents a 2.9-point pullback from the prior month. The brand remains classified as stable overall, but the monthly pattern shows an August peak followed by a September retreat.

The brand recorded 234 valid recommendations out of 645 qualified observations, with 325 total mentions. Sentiment is strongly positive at 0.8492, with 276 positive mentions, 49 neutral mentions, and zero negative mentions across the tracked set. The brand's strongest platform signal comes from Gemini, where it holds 41.6% valid recommendation coverage and a 5.6% rank-one rate, the highest rank-one rate the brand achieves on any platform.

The clearest gap is at the top of the recommendation structure. Tower 28's presence rate of 50.4% is the third-highest in the category, but its rank-one rate of 2.5% places it in the middle of the tracked set rather than near the top. The brand is being surfaced in half of all qualified observations, yet it is rarely the first recommendation AI systems offer.

The weakest platform signal is ChatGPT, where Tower 28 records zero rank-one recommendations across 80 observations despite a 60.0% presence rate. The brand appears frequently in ChatGPT answers but is never placed first, suggesting a recommendation conversion problem specific to that surface.

What Tower 28 Is Winning

Questions This Section Answers

• What is Tower 28's clearest strength in clean makeup AI recommendations?
• Where does Tower 28 show its strongest platform-level performance?

Tower 28's clearest win is its absence of negative framing. The brand recorded zero negative mentions across all tracked platforms in September 2026, a distinction shared with only a few competitors in the tracked set. Its net sentiment score of 0.8492 reflects consistently positive framing when the brand is mentioned.

The brand also holds a strong presence position. Its raw mention presence rate of 50.4% is the third-highest in the category, behind only Rare Beauty at 66.0% and e.l.f. Cosmetics at 61.4%. Tower 28 is being surfaced in half of all qualified AI observations, which provides a substantial base for recommendation conversion.

Gemini is a meaningful pocket of strength. Tower 28 achieves 41.6% valid recommendation coverage on Gemini, its highest coverage rate across all tracked platforms, with a 5.6% rank-one rate and a 17.98% top-three rate. The brand also holds a 56.2% positive visibility rate on that platform, its strongest positive framing anywhere in the tracked set.

Where Tower 28 Has the Clearest AI Visibility Gaps

Questions This Section Answers

• Why is Tower 28's presence-to-rank-one conversion the weakest among the top five brands?
• What is the clearest platform-specific weakness in Tower 28's recommendation profile?

Tower 28's central gap is the conversion of presence into top placement. The brand is mentioned in 50.4% of qualified observations but appears first in only 2.5% of them. By comparison, e.l.f. Cosmetics holds a 61.4% presence rate and converts it into a 14.3% rank-one rate, while Rare Beauty holds a 66.0% presence rate and achieves a 6.0% rank-one rate. Tower 28's presence-to-rank-one conversion is the weakest among the top five brands by coverage.

The ChatGPT gap is the clearest platform-specific weakness. Across 80 ChatGPT observations, Tower 28 is present in 60.0% of answers but records zero rank-one recommendations and only an 8.75% top-three rate. The brand is being surfaced consistently on ChatGPT but is never the first recommendation, indicating that other brands are capturing the top position in the answers where Tower 28 appears.

The brand's average recommended rank of 3.66 also signals a placement problem. When Tower 28 is recommended, it tends to appear lower in the shortlist than its coverage rate would suggest. The category leaders hold average recommended ranks near 2.6, while Tower 28 sits closer to the middle of the recommendation list.

Biggest Opportunity

Questions This Section Answers

• What is Tower 28's biggest opportunity for improving its AI recommendation position?

Tower 28's clearest opportunity is converting its high reference rate into stronger top-three and rank-one placement on ChatGPT. The brand is present in 60.0% of ChatGPT answers but never appears first, and its top-three rate on that platform is less than half its overall average. This is the single largest gap between presence and recommendation conversion in the brand's profile.

The path forward is to identify which ChatGPT prompts surface Tower 28 as context rather than as a chosen option, then build the owned answer layer and citation architecture that positions the brand as the first recommendation in those high-intent discovery moments. The brand already holds the presence base; the missing piece is the recommendation-stage evidence that moves it from mentioned to selected.

Competitive Landscape

Questions This Section Answers

• How does Tower 28's recommendation placement compare with the category leaders?

e.l.f. Cosmetics and Rare Beauty hold the strongest recommendation-stage positions in the clean makeup category, with Tower 28 sitting third but trailing the leaders by a meaningful margin on top-three and rank-one placement.

Average recommended rank covers rank-eligible recommendations only.

The table shows Tower 28 holding third place on coverage but trailing e.l.f. Cosmetics and Rare Beauty by roughly 10 points on top-three rate and by a wider margin on rank-one placement. The brand's sentiment is strong, but its recommendation position sits below the leading pair.

Prompt Evidence

Questions This Section Answers

• What do the platform-specific prompts reveal about how Tower 28 is being surfaced versus selected?

Gemini / Best Clean Makeup Brands Discovery & Evaluation Prompt: "What is the best blush on the market?" Result: Tower 28 appears in the recommendation shortlist with its strongest platform-level rank-one rate of 5.6%, indicating Gemini surfaces the brand as a viable option in product-specific discovery prompts.

ChatGPT / Best Clean Makeup Brands Discovery & Evaluation Prompt: "What is the most popular makeup brand?" Result: Tower 28 is present in 60.0% of ChatGPT answers but records zero rank-one recommendations, suggesting the brand appears as context or comparison rather than as the selected first option.

Perplexity / Best Clean Makeup Brands Discovery & Evaluation Prompt: "best setting spray" Result: Tower 28 achieves 36.1% valid recommendation coverage on Perplexity but a 0.0% rank-one rate, indicating the brand is shortlisted regularly yet never placed at the top of the answer.

What CiteWorks Studio Would Do Next

Phase 1: AI Market Discovery Audit Map which high-intent prompts surface Tower 28 as a mention versus a recommendation, with particular focus on ChatGPT where the presence-to-rank-one gap is widest.

Phase 2: Recommendation Readiness Plan Identify the product categories and prompt types where Tower 28 is shortlisted but not selected first, then prioritize the discovery moments with the highest commercial value.

Phase 3: Owned Answer Layer Buildout Develop product-level content that gives AI systems clear, citable reasons to place Tower 28 first in response to specific clean makeup discovery prompts.

Phase 4: Citation / Authority Layer Development Strengthen the external source footprint that supports Tower 28's recommendation eligibility, focusing on the evidence layer AI systems appear to synthesize when building shortlists.

Phase 5: Monthly AI Visibility and Recommendation Tracking Track whether the presence-to-rank-one conversion gap narrows across ChatGPT and other platforms as the owned answer and citation layers mature.

Why This Matters

Questions This Section Answers

• Why does being mentioned without being ranked first matter for Tower 28's commercial outcomes?

AI systems are surfacing Tower 28 in half of all clean makeup discovery answers, but they are rarely choosing the brand first. In buyer-choice terms, the brand is on the shortlist without winning the decision moment. That distinction matters because the first recommendation in an AI answer carries disproportionate influence over which brand a shopper explores next.

The next move is not broader visibility. Tower 28 already has that. The next move is targeted correction of the prompt, page, and citation layers that determine whether the brand is mentioned as an option or recommended as the answer.

Core Metrics

Sentiment Score

Sentiment Score = (positive mentions × 1 + neutral mentions × 0 + negative mentions × -1) / total mentions

For Tower 28, the calculation is (276 × 1 + 49 × 0 + 0 × -1) / 325, producing a net sentiment score of 0.8492.

This matters because unclassified mention counts are misleading. A brand can appear frequently in AI answers while being framed negatively or as a cautionary example, and raw mention volume would hide that distinction. Share of voice is a diagnostic metric, not a business KPI. A positive recommendation, neutral reference, cautionary mention, and competitor-displaced mention are not equal, and counting all mentions as wins is bad measurement. Classified sentiment is required before interpreting AI visibility, because it reveals whether a brand is being recommended, referenced, or warned against.

Sentiment by Platform

Methodology

1. This report is a benchmark-based analysis of Tower 28's AI recommendation visibility in the clean makeup category, produced from the LLM Authority Index AI Market Discovery Index and CiteWorks Studio interpretation of that public benchmark data. It is not a client implementation case study.
2. The reporting window is September 2026, with comparison references to July 2026 and August 2026 baseline and intermediate months.
3. Six canonical AI and search surface families were tracked: ChatGPT, Copilot, Gemini, Perplexity, AI Overviews, and AI Mode.
4. The benchmark began with 800 prompt-surface observations in September 2026, of which 699 were relevant to the vertical and 645 qualified for the public denominator after two qualification stages.
5. The competitor universe includes 10 tracked brands: Beautycounter, e.l.f. Cosmetics, Glossier, ILIA Beauty, Kosas, Milk Makeup, Rare Beauty, Tarte Cosmetics, Thrive Causemetics, and Tower 28.
6. All 645 qualified observations in September 2026 fell into the Brand Recommendation buyer-intent class. The public benchmark does not yet contain qualified observations in pricing and value or multi-brand comparison classes.
7. Stage 0 extraction captured prompt-level observations including query, AI surface, answer, brand outcome, recommendation placement, sentiment, and citations where exposed.
8. A mention is defined as any qualified observation where the brand appears in the AI answer in any context, whether recommended, referenced, or compared.
9. A valid recommendation is defined as a qualified observation where the brand appears in a recommendation shortlist. Negative, neutral, cautionary, comparison-anchor, and listed-only mentions are not counted as valid recommendations.
10. Brand-level percentages use the 645 qualified observations as the public denominator, not the raw collection of 800 prompts.
11. The public benchmark does not measure market share, sales attribution, organic-search ranking performance, social media volume, or private channels. Source presence is evidence about the information environment, not proof of causation.
12. Limitations: Small-count brands carry more month-to-month variance. Movement analysis identifies changes worth investigating, not evidence of specific causes. The public benchmark measures which brand gets recommended, not how AI systems frame trade-offs on price, value, or direct comparison.

Get Your AI Visibility Audit

The public benchmark shows where Tower 28 is winning and losing in AI-driven discovery. A company-level audit goes deeper, mapping the specific prompts, competitor displacement patterns, and evidence sources that determine whether the brand is mentioned or recommended first. That is where the story behind the movement becomes actionable.