How AI Search Is Recommending Roofing Companies: Monthly Trends

Executive Summary

The roofing category held its shape in October 2026, with GAF retaining the coverage lead at 53.6% valid recommendation coverage on 534 qualified observations and no brand exceeding normal month-to-month variation. The gap to second-placed Owens Corning, at 51.7%, is 1.9 points, and the gap to third-placed CertainTeed, at 49.4%, is 4.2 points. The distinguishing feature of October is not a leadership change: the top of the category narrowed sharply in September before easing back toward its July spread, while GAF held the lead throughout.

TAMKO's third consecutive monthly increase is the clearest trend-level signal this month, rising from 10.6% in September 2026 to 12.4% in October 2026 on 66 valid recommendations; the cumulative move remains inside normal variation. Malarkey posted the steepest single-month decline among brands with meaningful coverage, falling from 42.7% in September 2026 to 37.8% in October 2026 on 202 valid recommendations. Neither move was flagged as significant.

Taken against the full series, GAF has moved from 48.4% in July 2026 to 53.6% in October 2026, a gain of 5.2 points, and Owens Corning from 47.6% to 51.7%, a gain of 4.1 points. CertainTeed has moved from 44.1% to 49.4%, and Malarkey from 35.1% to 37.8%. Seven of the ten tracked brands sit above their July 2026 readings, and the category's top three brands are separated by 4.2 points in October 2026, little changed from 4.3 points in July 2026 after narrowing to 2.4 points in September 2026.

Atlas Roofing is the only brand carrying a significant baseline movement, up 5.8 points from 20.4% in July 2026 to 26.2% in October 2026, though its October reading eased 1.3 points from September 2026. DECRA is the only significant decliner, down 0.9 points from 0.9% in July 2026 to zero.

Each monthly run begins with 800 prompt-surface observations (504 unique questions) across the benchmark's defined AI/search surface universe. Of those, 800 mentioned a tracked brand or competitor; 785 were relevant and 15 were irrelevant. The public metrics use the 534 observations that survive both qualification stages, against 533 in July 2026 (455 unique questions; 783 relevant, 17 irrelevant). September 2026 produced 539 qualified observations from 492 unique questions.

Key Findings

AI Response Inconsistency Alerts

Questions This Section Answers

• Which roofing product claims did AI platforms contradict each other on?
• What caused the inconsistencies between platforms citing the same source page?

Six critical or high-severity factual inconsistencies were detected across five AI platforms, spanning four companies in the roofing category. Each involves two platforms returning incompatible claims in response to the same buyer question, and several trace back to cited source pages that contradict one another.

Malarkey

AI platforms provided conflicting information about the impact rating of the Malarkey Vista shingle line. When asked "Is Malarkey a good roofing product?", Google AI Mode stated that Vista has a Class 3 impact rating, citing Malarkey's own product pages alongside a third-party review and a comparison video. Copilot stated that Vista has a Class 4 impact rating for the same question, citing a shingles review site, a brand review page, and the same third-party review. The flagged sources split across both sides: Malarkey's own site states "Impact Resistance 4 Up to Class 4 hail impact rating (highest rating possible)," while a distributor page describes "a shingle family with standard-issue Class 3 and Class 4 impact ratings." A shingle line cannot be both Class 3 and Class 4 impact rated, as these are mutually exclusive standardized classifications.

A second conflict on the same topic surfaced between ChatGPT and Google AI Mode. When asked "Is Malarkey a good roofing product?", ChatGPT stated that Vista carries a Class 4 impact rating, citing Malarkey's sustainable roofing page. Google AI Mode stated that Vista has a Class 3 impact rating, citing Malarkey's homepage, a third-party review, and a comparison video. The flagged sources again diverge: a third-party review states "Even their entry-level Highlander line uses this technology and provides a Class 3 Impact Rating," while Malarkey's own site states "Impact Resistance 4 Up to Class 4 hail impact rating (highest rating possible)."

TAMKO

AI platforms provided conflicting information about the warranty length on TAMKO Heritage shingles. When asked "Is Tamko Heritage a good shingle?", ChatGPT stated that Heritage carries a limited lifetime warranty for qualifying single-family applications, citing TAMKO's Heritage page and a retailer product listing. Google AI Mode stated that the standard Heritage line carries a 30-year limited warranty, citing two roofing forum threads and a buying guide. The flagged sources include a TAMKO page noting "Limited Lifetime only applies to single-family structures," which supports the lifetime position, alongside a contractor page referencing lifetime limited coverage for a different brand's product line.

A second warranty conflict appeared between Google AI Overviews and ChatGPT. When asked "Are TAMKO Heritage shingles any good?", Google AI Overviews stated that "the flagship Heritage shingle typically carries a 30-year limited warranty, falling short of lifetime limited warranties," citing a review video, TAMKO's Heritage page, and a roofing contractor blog. ChatGPT stated that current Heritage shingles have a Limited Lifetime Warranty for qualifying single-family installations, citing a trade publication, TAMKO's Heritage page, and a consumer product review. Both responses drew on TAMKO's own Heritage page, which states "Limited Lifetime only applies to single-family structures," indicating the same source was read to support incompatible conclusions.

CertainTeed

AI platforms provided conflicting pricing information about the CertainTeed Landmark PRO shingle. When asked "Are CertainTeed shingles more expensive?", Gemini stated a material price range of $105 to $130 per square, citing a 2026 brand guide, a market comparison page, and a regional price guide. Google AI Overviews stated a cost of $150 to $200 per square, citing a brand comparison blog, a regional roofing comparison, and a contractor product page. The two ranges do not overlap, so both cannot be correct for the same product's material cost.

Owens Corning

AI platforms provided conflicting availability information about the Desert Tan color in the Owens Corning TruDefinition Duration line. When asked "Are desert tan shingles discontinued?", ChatGPT stated that Desert Tan is still available in certain Owens Corning product lines, including some Oakridge, Duration, Duration FLEX, and Supreme offerings, depending on region and distributor, citing a contractor blog, Owens Corning's shingle widget, and a product catalog page. Google AI Overviews stated that Owens Corning's TruDefinition Duration architectural shingles in Desert Tan have been discontinued, citing the same contractor blog, a roofing forum thread, and a retailer listing. The flagged source is unambiguous on one side: a contractor page states "Discontinued: Owens Corning TruDefinition® Duration® Shingle in Desert Tan," while a separate page notes that "popular shingles like CertainTeed XT™ 25 or Owens Corning Desert Tan have been quietly discontinued."

Benchmark Context

Questions This Section Answers

• What is the qualified benchmark set, and why does it differ from the raw collection universe?
• How did the qualified observation count and surface families change between July and October 2026?

Research-Scope Funnel

The table below separates the raw collection universe from the qualified analysis set. Brand-level recommendation percentages are calculated within the qualified benchmark set, not the full prompt pool.

The public denominator is effectively unchanged, from 533 qualified observations in July 2026 to 534 in October 2026, though the two intervening months moved materially: August 2026 expanded to 610 with a seventh surface family (Grok) present, and September 2026 returned to 539 on six surfaces. October 2026 holds the same six canonical surface families observed in July 2026 and September 2026: ChatGPT, Copilot, Gemini, AI Overviews, AI Mode, and Perplexity.

Benchmark-Level Metrics

The intermediate months matter for reading the aggregate picture. Recommendation-shaped answer share peaked at 29.3% in August 2026, fell to 26.2% in September 2026, and returned to 29.0% in October 2026. Valid recommendation shortlist share peaked at 57.7% in September 2026 and settled at 55.4% in October 2026, still 7.4 points above the July 2026 baseline.

AI Recommendation Trend

Questions This Section Answers

• Who leads roofing recommendation coverage, and by how much?
• Which roofing brands gained the most coverage across the series?

A Stable Leader Sits Atop a Compressed Top Three

GAF holds the coverage lead for the fourth consecutive month, and the gap to Owens Corning widened from 0.1 points in September 2026 to 1.9 points in October 2026, driven by a 2.7-point decline in Owens Corning's coverage rather than any GAF advance. GAF's own reading fell 0.9 points over the same span. The category is best described as stable, with the top three positions settling back close to where they started the series.

Two brand entries in the tracked set, CertainTeed and Malarkey, carry zero coverage across every month of the series and rank below DECRA on current reading. No brand exceeded normal month-to-month variation in October 2026. The category-level change over the series came from the combination of several smaller movements, led by Atlas Roofing's significant gain and the broad upward drift across the top five.

What Changed This Month

Questions This Section Answers

• Why did Malarkey post the sharpest single-month decline?
• Why is Atlas Roofing's significant coverage gain not translating into top-three placement?
• What connects the September-to-October easing across the top four brands?

GAF: A Stable Lead With a Distinctive Rank-One Profile

GAF's valid recommendation coverage moved from 48.4% in July 2026 (258 valid recommendations on 533 observations) to 53.6% in October 2026 (286 valid recommendations on 534 observations), a gain of 5.2 points. The month-over-month direction was down 0.9 points from 54.5% in September 2026.

Placement tells a different story. GAF's top-three rate rose from 42.8% in July 2026 to 50.6% in October 2026, a 7.8-point gain, and its rank-one rate held essentially flat at 28.8% against 28.7%. Its October 2026 rank-one count is 154 recommendations, the highest in the category.

GAF is the only brand in the tracked set whose rank-one rate stands roughly four times its nearest peer. Raw mention presence sits at 98.7%, meaning nearly every qualified response names the brand. Highest-priority diagnostic: which prompt types and surfaces produce GAF's rank-one positions, and whether that concentration rests on a narrow set of evidence sources.

Malarkey: The Sharpest Single-Month Decline Among Covered Brands

Malarkey's valid recommendation coverage fell from 42.7% in September 2026 to 37.8% in October 2026 on 202 valid recommendations, a decline of 4.9 points that falls within normal month-to-month variation. Against the July 2026 baseline of 35.1% on 187 valid recommendations, the October 2026 reading remains 2.7 points higher.

The decline is not a visibility decline. Raw mention presence moved from 50.3% in July 2026 to 47.2% in October 2026, a 3.1-point easing that is also inside normal variation. Net sentiment held at 0.9 across both months, the highest in the tracked set.

The distinction worth noticing is between recommendation-set inclusion and placement. Malarkey's top-three rate stands at 3.9% and its rank-one rate at 0.6%, both below their July 2026 readings, meaning the brand is named in roughly half of qualified responses but placed near the front in a small fraction of them. The two zero-coverage brand entries at the bottom of the trend table share the Malarkey name string and should not be read as a separate commercial entity. Highest-priority diagnostic: which prompts dropped Malarkey from the valid recommendation set between September 2026 and October 2026, and whether the Vista impact-rating inconsistency identified above appears in those same recommendation contexts.

Atlas Roofing: The Series' Only Significant Riser Eases Back

Atlas Roofing moved from 20.4% in July 2026 (109 valid recommendations) to 26.2% in October 2026 (140 valid recommendations), a gain of 5.8 points flagged as significant and the largest baseline-to-current gain in the category. The brand's month-over-month direction was down 1.3 points from September 2026.

The composition is split. Atlas Roofing's top-three rate moved from 0.8% in July 2026 to 0.9% in October 2026 on 5 top-three recommendations, and its rank-one rate from zero to 0.4% on 2 rank-one recommendations. Raw mention presence eased 1.4 points to 35.6%.

Atlas Roofing's coverage gain has not converted into placement. Roughly 140 valid recommendations place the brand in the set, and 5 place it in the top three. Its net sentiment score of 0.8 is among the highest in the tracked set. Highest-priority diagnostic: which prompts added Atlas Roofing to the recommendation set without moving it into top-three position, and what distinguishes the small handful of prompts where it does place.

TAMKO: A Third Consecutive Monthly Gain Inside Normal Variation

TAMKO rose from 10.1% in July 2026 (54 valid recommendations) to 12.4% in October 2026 (66 valid recommendations), a gain of 2.3 points. The month-over-month move was up 1.8 points from 10.6% in September 2026, the brand's third consecutive monthly increase, with the cumulative change still within normal variation.

Presence and placement have moved independently. Raw mention presence edged from 31.0% in July 2026 to 31.8% in October 2026, a 0.8-point change. The top-three rate held flat at 1.7%, and average recommended rank eased from 4.76 in July 2026 to 5.23 in October 2026, meaning the brand is being added to more recommendation sets while its typical position within them slips.

TAMKO's coverage gain is arriving through set inclusion rather than placement. Net sentiment eased from 0.5 to 0.4 across the series. Highest-priority diagnostic: whether the three-month coverage climb is stable enough to persist, and which prompts produced the additional valid recommendations while average rank was declining.

The Coverage Compression at the Top

GAF, Owens Corning, CertainTeed, and Malarkey all posted their series high in September 2026 and eased in October 2026: GAF down 0.9 points, Owens Corning down 2.7 points, CertainTeed down 2.7 points, and Malarkey down 4.9 points. None of those moves reached significance.

The top four brands span a 15.8-point band in October 2026, close to July 2026's 13.3-point spread after narrowing to an 11.8-point band in September 2026, even as all four eased from their September highs by similar magnitudes. Category-level notable gaps widened across the series between Atlas Roofing and each of DECRA, Erie Home, and Power Home Remodeling, by 6.7, 6.2, and 6.2 points respectively, none of them widening every month. That pattern reflects shifts in the tracked set's relative standing rather than a single brand's problem. Highest-priority diagnostic: whether the September 2026 to October 2026 easing across the top four reflects a shared prompt-mix or surface effect rather than four independent brand movements.

Buyer-Intent Interpretation

Questions This Section Answers

• What kinds of buyer questions does the roofing benchmark actually capture?
• What can't the benchmark measure about pricing and brand comparison questions?

Every qualified observation in October 2026 fell into the Brand Recommendation cluster. No qualified observations landed in Pricing & Value or Multi-Brand Comparison, so the public benchmark can answer which brand AI systems name when a buyer asks for a recommendation, but not how those systems position brands on cost, value, or head-to-head comparison.

The commercial gap is significant given the inconsistency data reported above. Buyers asking whether CertainTeed shingles are more expensive received non-overlapping price ranges from Gemini and Google AI Overviews, and buyers asking whether Desert Tan shingles are discontinued received incompatible availability answers from ChatGPT and Google AI Overviews. Those questions fall outside the qualified benchmark set, so their effect on recommendation coverage cannot be measured from this data.

Brand Opportunity Summary

Questions This Section Answers

• What is the highest-priority diagnostic for each tracked roofing brand?

The benchmark identifies where attention is warranted; a company-level analysis is needed to explain why.

Evidence Behind the Benchmark

Questions This Section Answers

• What data does the benchmark capture, and what does it not establish?

The aggregate metrics are built from prompt-level observations capturing the query, the AI surface, the recommendation outcome, the ranking position, sentiment, and citations where exposed. Company-level analysis can go deeper into prompt, competitor, surface, and evidence-source patterns. Source presence in an AI response is not automatically treated as proof of causation; it is a signal that the source shaped the answer, not that it caused the recommendation outcome. The inconsistency data reported above illustrates why that distinction matters: in several cases, two platforms cited the same source page and reached opposite conclusions about the same product attribute.

About This Benchmark

This report is part of the LLM Authority Index AI Visibility Market Discovery research program.

• AI Visibility Industry Market
• AI Visibility Industry Market Methodology
• AI Visibility Industry Market Metrics
• AI Visibility Industry Market Standards

Report-Specific Interpretation Notes

• Small-count movement: Power Home Remodeling's 2.4% coverage rests on 13 valid recommendations, Erie Home's 2.2% on 12, and DECRA's zero on 2 present observations. These figures are directionally indicative rather than robust.
• Qualified denominator versus raw collection: percentages are calculated against the qualified benchmark set of 534 observations in October 2026, not the 800 raw prompt-surface pairs, so shifts in the qualified set can move rates even when the raw collection is stable.
• Directional analysis: month-over-month movement identifies changes worth investigating. It does not by itself establish the cause of those changes. The September 2026 to October 2026 easing across the top four brands may reflect a shared prompt-mix or surface effect, and the data alone cannot distinguish among possible explanations.

Next Step

The Public Benchmark Shows Where a Brand Is Winning or Losing. A Company-Level Audit Shows Why.

The aggregate percentages in this report do not reveal which high-intent prompts are won or lost, which competitor takes the recommendation when a brand loses, what attributes AI systems associate with each roofing option, or which external sources shape those answers. The inconsistency findings sharpen the question further: the same manufacturer page was cited by two platforms that returned incompatible claims about a shingle's impact rating and another brand's warranty length, and only a prompt-level audit can show where those contradictions surface in a buyer's path.

A company-specific AI visibility audit maps those prompt, surface, competitor, ranking, sentiment, and evidence-source patterns into a prioritized visibility strategy. It identifies which gaps are addressable through owned content, which require third-party source cultivation, and which surfaces deserve attention first.

Request an AI visibility audit