Study Parameters

Executive Summary

We wanted to find out one thing: if a brand owns a software category on Google, does it also win in AI search? So we took the domain with the highest Ahrefs traffic share in each of 134 B2B software categories and checked where it ranked on Google versus where four AI models ranked it.

The answer: sometimes, but it's messy.

Only 35.1% of categories lined up cleanly (the AI rank was within 2 spots of Google, and at least 3 models mentioned the brand). Almost exactly the same share, 35.8%, went the other way. In those, the AI models either ranked the topical authority way lower than Google or skipped it completely. The rest, 29.1%, landed somewhere in between. The link between Google rank and average AI rank is weak (Pearson r = 0.145, p = 0.096), and it doesn't clear the bar for statistical significance.

Here's the part that stands out. When AI disagrees with Google, it almost always ranks the topical authority brand lower, not higher. On average it drops 3.65 positions. That tells us AI models lean on how famous a brand is, not how much great content it has published. The biggest jump up belongs to lastpass.com (Google rank 19, average AI rank 5.75). The biggest drop belongs to transcribe.com (Google rank 10, average AI rank 31.5). Same cause in both cases: how well-known the brand is outside its little SEO keyword bubble.

If you're a B2B SaaS brand that poured money into content-led SEO, here's the warning. Owning a Google keyword cluster does not buy you a seat in AI search. The brands AI surfaces with confidence are the ones with broad web presence, press, and a long review history. Not the ones with the highest Ahrefs traffic share.

The Findings at a Glance

One Thing to Keep in Mind

AI search is highly probabilistic. There's a real element of randomness baked into this channel, and the same prompt can return different answers on different days. We've done our best to strip out as much of that randomness as possible and control the variables we could. But there are still plenty of factors at play, so some of these conclusions won't hold across the board.

Here's the honest version of the thesis: these findings are about software companies specifically, looking at topical authority between Google and AI search. We're not claiming this is the most scientific research ever done on the topic. It's a clear, practical read on a pattern we kept seeing.

Methodology

How we found the topical authority brand

The way we define topical authority here is simple: a company's traffic share within its category. We think traffic share, as a percentage, is a really good proxy for how authoritative a brand is around a given topic cluster. If a brand captures a big slice of all the search traffic for a category's keywords, it has done the most to own that topic in Google's eyes.

Here's how we measured it:

1. For each category, we took the seed keyword (say, "email marketing"), dropped it into the phrase match report, and exported every keyword that contained that term, usually thousands of them.

2. We added that whole pool back into Ahrefs in bulk and looked at the traffic share by domain across the entire set.

3. Then we took the top-ranking software company in that pool and recorded its traffic share

4.  We excluded anything that wasn't a software company itself: Reddit, Quora, review aggregators, and any other non-software sites.

5. So a brand's traffic share is the percentage of all the category-keyword traffic that one domain captures, measured against everyone else competing for those same terms. 

We ran the exact same process across every category and every brand.

How we collected Google ranks

We didn't do this by hand. We pulled the rankings through the Ahrefs API, set to the US region and English. We recorded positions 1 through 20. Anything outside the top 20 got marked 21+. If a domain showed up more than once, we only counted its best position. AI Overviews were noted on the side and never counted as an organic position.

How we collected AI ranks

Through API automation across ChatGPT (GPT-4o), Google Gemini, Anthropic Claude, and Perplexity AI. One response per model per category prompt. We pulled brands in the order they first appeared. "Not mentioned" got recorded as NM. The average AI rank ignores NM values.

One thing worth acknowledging: because we ran these through the API, the results may look a little different from what someone gets searching the same prompt in their own AI window. A real user's results are shaped by their chat history and everything the platform already knows about them, which we don't control for here. That's fine — we're not claiming the API output is the one true answer. It still gives a solid read on visibility and ranking inside AI search.

How we classified consistency

Module 1: Topical Authority vs. Google Rank

The topical authority domain landed in Google's top 3 in 53% of categories, and in the top 10 in 96.3% of them. The median Google rank was position 3. So the Ahrefs traffic share method holds up: the domains it flags as topical leaders really are prominent in organic search.

Only 5 of the 134 domains missed the top 10, and all 5 still made the top 20. Not a single topical authority domain was missing from the first two pages of Google. So the gap this study is chasing isn't between topical authority and Google. It's between Google and AI search.

Figure 1: Google rank distribution of topical authority domains across 134 B2B software categories.

What this actually means

Picture the Ahrefs topical authority domain as the brand that has written the most, and the most-read, content about a software category. This module shows Google almost always puts that brand near the top, which is exactly what you'd expect. The real question is whether AI assistants do the same. That's what the rest of the report digs into.

Module 2: Topical Authority vs. LLM Rank

Across the four models, the topical authority domain's average rank slid from Google's baseline of 4.04 down to an average AI rank of 7.69. That's a 3.65-position drop. Claude treated these brands best (mean rank 6.73). Gemini treated them worst (mean rank 10.46). Perplexity stuck closest to Google's order.

The per-model correlation with Google tells the same story from another angle. Only Perplexity has a real, significant link to Google (r = 0.299, p = 0.0004). ChatGPT (r = 0.030) and Gemini (r = 0.056) are basically at zero, which means their rankings are driven by something that has almost nothing to do with topical search authority. Claude sits in the middle (r = 0.154).

Figure 2: Mean rank of the topical authority domain by model. Delta shown versus the Google baseline.

What this actually means

Search Google for 'best CRM software' and the company that has published the most CRM content will probably sit near the top. Ask an AI the same question and you get a different answer, and how different depends a lot on which AI you ask. Perplexity acts the most like Google. Gemini acts the least like it. That matters because if you built your visibility through SEO content, you can't assume that same work carries over to AI search.

Module 3: Google Rank vs. LLM Rank (Correlation)

The scatter plot below puts each domain's Google rank on the x-axis and its average AI rank on the y-axis. The red dashed diagonal is perfect agreement. Points above the line mean AI ranks the domain lower than Google. Points below mean AI ranks it higher.

The cluster of points sitting above the diagonal is the whole story in one picture: for most domains, AI ranks the topical authority lower than Google does. The trend line (amber) tilts up but barely, which confirms the weak correlation (r = 0.145, p = 0.096). The two big exceptions, adobe.com and zoom.com sitting way below the line, come down to one thing: those brands are everywhere in training data. They're among the most-referenced software brands on the internet.

Figure 3: Scatter of Google rank (x) vs. average AI rank (y). Diagonal = perfect alignment. Points above = AI ranks lower. Notable outliers labeled.

What this actually means

This chart is the statistical core of the report. If AI just copied Google, every dot would sit on the red diagonal. The fact that most dots float above it tells you AI is steadily ranking content-authority leaders lower than Google does. And the almost-flat trend line tells you that where a brand sits on Google is a bad predictor of where it shows up in AI answers. The two are measuring different things.

Module 4: Rank Delta Analysis

The chart below shows the full spread of (average AI rank minus Google rank) across all 134 domains. It leans hard to the right, which confirms the typical gap runs one way: AI under-ranks topical authority domains compared to Google, not the other way around.

Figure 4: Distribution of rank delta. Green = AI ranks higher than Google. Grey = consistent zone. Red = AI ranks lower.

Domains where AI ranks higher than Google (AI boost, top 10)

Every name on this list is a globally recognizable software brand. These companies show up constantly in pre-training data, product reviews, news, and general web text. AI has learned a strong tie between the brand name and its category, completely separate from where it ranks on Google today. Lastpass is the wildest case: 19th on Google, yet placed in the top 6 by every single model.

Domains where AI ranks lower than Google (AI miss, top 10)

The AI miss list is full of niche SaaS players that built their Google presence by pumping out a high volume of content inside a narrow keyword cluster, not by building broad brand recognition. These domains are SEO-rich but invisible to AI. In several cases, a dominant incumbent (Zoom for VoIP, Salesforce for CRM) swallows the category's traffic leader in AI answers no matter who holds the Ahrefs traffic share spot.

There are two key insights here:

• The AI miss pattern clusters in niche verticals where the topical authority domain built its Google presence on content volume in a narrow cluster, not broad brand recognition. These domains are SEO-rich but invisible to AI.

• The AI boost pattern strongly favors legacy tech brands (Adobe, Zoom, Salesforce, Zendesk, HubSpot, Lastpass) whose names show up constantly in pre-training data, no matter where they sit on Google today.

What this actually means

Think of it this way. Google rewards the brand that wrote the most relevant content. AI rewards the brand it has heard the most about. For famous software companies, those are often two different brands. A small SaaS tool can own a Google keyword cluster with a smart content play, but if it never shows up in tech press, review sites, or user communities, AI doesn't know it exists. Flip it around: a brand like Adobe or Zoom is so baked into the public conversation about software that AI hands it over as the answer whether or not it ranks on page one of Google.

Module 5: Category-Type Patterns (Search Volume Tier Analysis)

We split the 132 categorized domains into three equal tiers of 44 by their Ahrefs USA search volume, straight from the master tracking sheet. Two domains with no SV recorded got dropped from this module.

Figure 5: Left, average Google and AI rank by search volume tier. Right, stacked consistency breakdown by tier.

Does search volume tier predict AI alignment?

The average delta climbs from High SV (+3.17) to Mid SV (+3.32) to Low SV (+4.62), which makes it look like low-volume niche categories take the biggest AI rank hit. But the average gets yanked around by extreme outliers in every tier. The median tells a cleaner story: High SV sits at just +0.75, Mid SV at +1.75, Low SV at +3.25. The High SV tier splits into two groups: a big cluster of well-aligned, well-known brands and a smaller group of badly divergent content-heavy players. The Low SV tier diverges more evenly and more moderately.

Here's the counterintuitive bit. The Low SV tier actually has the best consistent rate at 38.6%, versus 34.1% for both High and Mid SV. Low SV also has the lowest divergence rate at 29.5% versus 40.9% for High SV. High SV has the worst divergence of the three, driven by cases where a famous incumbent owns the AI answer no matter who holds the Ahrefs traffic share spot.

Traffic share vs. AI rank and Google rank, by tier

Figure 6: Scatter of traffic share % vs. average AI rank by tier. Negative slope = more traffic share predicts a better AI rank.

Figure 7: Scatter of traffic share % vs. Google rank by tier. Negative slope = more traffic share predicts a higher Google rank.

The correlations expose a real structural split across tiers. In Low SV categories, all three relationships are significant: traffic share predicts Google rank (r = -0.518), Google rank predicts AI rank (r = 0.487), and traffic share predicts AI rank (r = -0.380). A clean signal chain exists. In High SV categories, only traffic share to AI rank holds up (r = -0.308), while the Google-to-AI link vanishes completely. In Mid SV categories, only traffic share to Google rank is significant (r = -0.367).

So the thing driving AI visibility changes by category type. In niche, low-volume categories, owning the keyword cluster on Ahrefs pulls Google prominence and AI visibility along together. In high-volume categories, brand familiarity in training data overrides both, and no amount of content gets a lesser-known brand into AI answers.

What this actually means

Simplest way to frame it. In small, specialist software categories, if you own the Google results, there's a decent chance AI recommends you too. The internet doesn't have many competing voices about parking management software or farm tools, so the brand Google trusts is usually the one AI learned about as well. In big, crowded categories like password managers or website builders, that logic falls apart. There are so many well-known names in those spaces that AI defaults to whoever is most famous, not whoever did the best SEO. Owning a keyword cluster is not the same as owning the conversation.

Recommendations

Conclusion

Across 134 B2B software categories, topical authority on Google does not reliably carry over to AI search. The weak correlation (r = 0.145), the 35.8% divergence rate, and the average rank drop of +3.65 positions all point the same way: AI isn't simply mirroring the topical depth signals behind Google rankings. It leans on broader brand familiarity instead, the kind that comes from being referenced widely across the internet, not from owning a single keyword cluster.

The study also shows the mechanism changes by category type. In niche, low-volume verticals, there's a clean signal chain from traffic share to Google rank to AI rank, but the absolute AI positions are the worst because niche tools have barely any training data footprint. In high-volume, mature categories, brand ubiquity in training data overrides every search signal, and the Google-to-AI link disappears entirely. Both create their own problems for B2B SaaS brands.

The data shows clear correlations between traffic share, Google rank, and AI visibility, but it can't prove cause. Other factors, including brand age, domain authority, PR spend, and product review volume, almost certainly explain a big chunk of the variance. As AI search keeps growing as a discovery channel for software buyers, the gap between traditional SEO authority and AI visibility is going to become a bigger strategic fault line. The brands that win will be the ones that build both: topical depth for Google, and broad web presence for AI.

Appendix: Full Data Table (134 Domains, Sorted by Delta)

Sorted by rank delta, largest positive first. Positive delta = AI ranks lower than Google. Negative delta = AI ranks higher.