// Research note

AI Search 2026 - what 145 industries tell us about being recommended

The strongest correlation with being recommended by ChatGPT, Gemini, or Perplexity is still classic search-engine presence. In the OppAlerts study released March 2026, Search Engine Appearances posted a Spearman rho of +0.241 against LLM recommendation score across 145 industries. SEO did not get killed by AI. It got promoted into a more competitive layer where the same signals matter more, not less.

What OppAlerts actually did

The study is the largest public correlation analysis published on LLM ranking factors to date. The methodology block is the legitimacy anchor: 145 industries analysed, 1,595 buyer personas generated, and 105,000+ prompts sent to ChatGPT 5.4 to capture which domains the model recommends per persona. The team cross-referenced those recommendations against 1.1 billion Common Crawl pages, 5 billion Reddit posts and comments, 4 billion+ backlinks indexed, and 300 million Wikimedia entities. For each recommended domain they extracted 13 technical signals, normalised them within each industry, and ran Spearman rank correlations against the LLM recommendation score.

Translation for a business owner: this is the first dataset that lets you say with statistical confidence which signals move the needle on AI-mediated recommendations, sector by sector. The full study is at oppalerts.com/LLM-Ranking-Factors.

The top 5 signals across every industry

Classic SEO is the foundation, not the past. The Strong-tier signals (Spearman rho 0.20 or higher) across all 145 industries pooled are:

1. Search Engine Appearances - rho +0.241. How many of your sector's top-100 SERPs your domain shows up in.
2. Best Search Engine Rank - rho +0.238. Your best position across those SERPs.
3. SE Outbound Links - rho +0.230. Whether SERP results link out to your domain.
4. Backlink Count - rho +0.204. Raw count from the Common Crawl host-level web graph.
5. Backlink Authority - rho +0.200. Weighted authority of the linking domains.

PageRank lands just below (rho +0.194), then Wikidata associations (+0.120), then Reddit comments (+0.111). Read that order again. The top five are the things SEO has been doing for fifteen years. The data corroborates what Ahrefs already published: 38% of AI Overview citations come from organic top-10 results. The model is not inventing a new web. It is reading the same one, with the same weights, more strictly.

Why averages mislead

The cross-industry pooled numbers are a starting point, not an answer. Per-sector correlations are often more than twice as strong as the global average, and the dominant signal differs by sector:

• Boutique and budget hotel chains - dominant signal is Wikidata at rho 0.552.
• OTAs (online travel agencies) - dominant signal is Wikidata at rho 0.461.
• Short-term rentals (apartment rentals and multifamily) - dominant signal is Search Engine Appearances at rho 0.518.
• Residential real estate brokerages - dominant signal is Backlink Count at rho 0.559.
• Commercial real estate listing marketplaces - dominant signal is Wikidata at rho 0.559.
• Accounting software - dominant signal is Wikidata at rho 0.515.

Three sectors in that list collapse onto Wikidata as the dominant moat. Two collapse onto search appearances and backlinks. If you run a hotel and you build your AI visibility plan around backlinks because backlinks ranked fifth across all industries, you are optimising for the wrong signal. Tourism in particular is split: short-term rentals lean on search visibility, hotels lean on entity associations. The sector matters more than the average.

The "AI killed SEO" frame is wrong

The dataset does not support the headline. What it supports is harder to summarise but more useful: AI promoted SEO into a more competitive layer where the same signals matter, weighted differently per sector, with less tolerance for the bottom 90%.

Whitespark has been calling this the "30x harder" problem: where a classic SERP would show ten blue links and reward positions 1 through 10, an AI-generated answer surfaces one or two sources. The funnel is the same but the mouth is narrower. The signals that put you in the top 10 still apply. They just need to put you in the top 1 or 2.

What this means for a business owner this month

Three actions, derived directly from the data, that you can run before the end of the month.

1. Pull your "Search Engine Appearances" benchmark. Count how many of your sector's top 50 SERPs your domain currently appears in. If the count is zero, that is the gap, and it is the strongest cross-industry signal in the study (rho +0.241).
2. Audit whether your business has a Wikidata entry. If your sector is hotel, OTA, real estate, or accounting-services-adjacent, Wikidata is your dominant signal (rho between 0.45 and 0.56). If the entry does not exist, create one with verified references to your Google Business Profile, your website, and at least two third-party listings.
3. Choose one focused pass - schema markup audit, Google Business Profile completeness, or backlink baseline check. Pick the one that maps to your sector's dominant signal from the table above. Not the cross-industry average. Your sector.

Related research

• AI and Local SEO Workshop - Madeira, May 2026 - the workshop that produced this research push.
• Zyppy 32 AI Citation Factors - the 32-factor view that complements OppAlerts.
• Boutique Hotels and the Wikidata Entity Moat - sector deep-dive: hotels plus Wikidata.

References

• oppalerts.com/LLM-Ranking-Factors - the full study, March 2026.
• whitespark.ca - the "30x harder" framing for AI-mediated discovery.
• ahrefs.com - 38% of AI Overview citations come from organic top-10 results.