Semrush published a study on September 8, 2026 showing that AI assistants and Google's conversational search surface together sent 0.48% of all sessions to manufacturing and industrial websites between January and July 2026, a period in which generative summaries spread across 57% of the sector's tracked search volume.

In Short

A research team measured how often AI answers mention manufacturing brands, which websites those answers link to, and how many visits AI tools actually send to industrial sites. Almost half a percent of traffic came from AI, while the brands named most often in AI answers were rarely the ones being linked. The gap means visibility inside an AI answer and a visit to a website are now two separate things, measured in different places.

Three datasets, one seven-month window

The research, credited to Luke Harsel with contributor Nikola Manojlović, combines three separate Semrush datasets covering January through July 2026. According to the published methodology, each measures a different stage of discovery, and the company describes the combination as directional rather than definitive.

The first dataset is keyword and results-page data. Semrush took the top 20 manufacturing domains from its Trending Websites ranking, pulled the top 1,000 keywords from each, and retained only those keywords where at least two of those domains held a top ranking. That filter produced a set of 458 keywords, tracked through monthly snapshots. The measurement was weighted by search volume rather than keyword count, which means a single high-volume term moves the figure more than a cluster of rarely typed ones.

The second dataset is clickstream traffic, covering United States sessions to sites across ten categories: chemicals, civil engineering, construction and maintenance, equipment and supplies, machinery, manufacturing, plastics and polymers, renewable energy, transportation and logistics, and warehousing.

The third draws on the Semrush AI Visibility database for the United States manufacturing and industrial vertical, across four systems: ChatGPT, Gemini, Google AI Mode and Google AI Overviews. For each month the analysis captured the top 25 domains by mentions and the top 25 by citations, then tracked the aggregate. AI traffic to those domains was pulled from the company's Traffic and Market Toolkit over the same seven months.

Semrush sells AI visibility tracking, and the study closes by pointing readers toward those products. The figures below are vendor measurements of a market the vendor also serves.

Generative summaries reached 57% of tracked search volume

Across the 458-keyword set, AI Overviews climbed from 38% of search volume in January 2026 to 57% in July, a 19-point move in seven months.

The composition of that growth matters more than the headline. Semrush notes that the feature is no longer confined to broad informational questions such as what a transformer is. Terms that gained a summary during the window include the product query "pneumatic cylinder" and the part number "csla2gf", along with the branded search "National Instruments". A part number is about as far down the funnel as an industrial search gets, and a branded query is the one an incumbent supplier expects to own outright.

That expansion sets up the study's central tension. A results page carrying a synthesized answer produces fewer outbound clicks, which would imply falling organic traffic across the sector. The traffic data complicates that expectation.

Direct visits dominate, AI barely registers

Clickstream measurement of the ten industrial categories produced the following share of sessions between January and July 2026: direct at 56.65%, organic search at 22.28%, referral at 12.89%, email at 3.52%, paid search at 2.58%, organic social at 1.33%, AI assistants at 0.45%, paid social at 0.19%, display advertising at 0.06%, and AI Mode at 0.03%.

Direct and organic search together account for 78.93% of sessions, the "nearly 80%" the study cites. The two AI channels combine for 0.48%.

Within that stable-looking mix, one movement is consistent: direct traffic share is growing while organic search traffic declines. Demand for industrial goods and services has not visibly shifted, according to the study. The route people take to reach a supplier has.

That substitution is the finding with budget consequences. Organic search is measurable at the query level, attributable to a page, and defensible in a quarterly report. Direct traffic is a bucket that absorbs whatever analytics cannot classify, including sessions that began with a conversation inside a chatbot and ended with a brand name typed into a browser bar.

Semrush cites its own survey of more than 600 business-to-business professionals, which found buyers using AI tools to build vendor shortlists and inform purchases worth more than $1,000, then following up through a Google search or a direct visit to a brand named in the answer. PPC Land covered that survey in July 2026, where 519 United States respondents reported AI shaping 92% of vendor shortlists while brand recognition determined notice in an AI answer for only 7%.

Gaetano DiNardi is quoted in the study describing the resulting measurement problem as the "dark SEO funnel". Signals such as rising direct traffic or branded search volume may indicate AI visibility working, without any AI referral appearing in analytics at all.

Mentions and citations produce two different leaderboards

The study separates two metrics that are frequently collapsed into one. A mention is a brand name appearing in the text of an AI answer. A citation is a link the system offers as a source.

Of the 15 top domains on each list, only two appear on both: Vevor and Grainger.

According to figures Semrush circulated to press on September 9, 2026, Boeing recorded 771,000 mentions across the four systems, ahead of 3M at 743,000 and John Deere at 732,000. On Google AI Overviews alone, Boeing led with 286,000. Those totals do not appear in the text of the published study, which names the top-mentioned brands in the order 3M, John Deere and Boeing without attaching counts to them in machine-readable form. The ordering in the published prose and the ranking in the circulated figures therefore do not match, and this article reports both rather than resolving the difference.

What is not in dispute is the absence. Boeing does not appear among the most-cited sources at all.

The study's explanation is structural. Mentions reward recognition: a century-old name with a wide product range is relevant to an enormous number of prompts, and most of the top-mentioned brands predate OpenAI and Gemini by roughly a hundred years. Citations reward the page that answers the question, which is rarely a manufacturer's own site.

The cited list is a different kind of website

The 15 most-cited manufacturing and industrial domains, as published, are engineerfix.com, vevor.com, thermofisher.com, sigmaaldrich.com, made-in-china.com, rs-online.com, thebluebook.com, grainger.com, thomasnet.com, procore.com, rsc.org, protoolreviews.com, globalspec.com, carrier.com and energysage.com.

Semrush classifies them as a reference and editorial site, a manufacturing brand and retailer, two manufacturing brands, a business-to-business marketplace, an industrial distributor, a construction industry directory, another industrial distributor, an engineering directory, construction management software, a professional membership body that publishes chemistry journals, a power tools review publication, an engineering directory, a manufacturing brand, and an energy comparison marketplace.

Directories, marketplaces, distributors and reference publishers hold the majority. Manufacturers hold four places out of fifteen.

Domain strength splits the two lists as well. Semrush measured Authority Score, its proprietary trust metric built on organic traffic, backlinks and spam signals, across both groups. The top-mentioned brands scored consistently. The top-cited sources did not: engineerfix.com, the single most-cited domain in the sector, carries a score of 17, while made-in-china.com sits at 77, higher than any brand on the mentions list. The study reads that spread as evidence that site authority tracks mentions more closely than citations.

One site ChatGPT loves and Google barely sees

The clearest platform divergence in the dataset involves engineerfix.com, which dwarfs every other source in ChatGPT citations while failing to reach the top-cited list on Gemini, AI Mode or AI Overviews.

Its Google footprint is thin. Semrush data puts the site at roughly 3,700 ranking keywords with little estimated traffic, following a brief rise around spring 2022 that did not hold. Yet Traffic Analytics estimates close to 300,000 monthly organic visits.

The resolution sits in the referral mix. According to the study, DuckDuckGo, Bing and Yahoo account for more than 68% of traffic to the site, while Google contributes under 2%. A site invisible on one SERP can be a heavily indexed authority on others, and the model that grounds its answers in those other indexes inherits that view.

For anyone tracking citation share, the operational consequence is that a strategy producing results on one assistant carries no guarantee on another, because the retrieval layers underneath them are not drawing from the same web.

Citations do not convert into visits at a predictable rate

Matching AI traffic against mention and citation counts, the study finds no clean correlation in either direction.

On the mentions side, vevor.com and grainger.com draw AI traffic that rivals or exceeds what their mention volumes would suggest, and siemens.com pulls more AI traffic than its mention count implies. On the citation side, engineerfix.com leads the sector in citations and collects only a fraction of the AI referrals that lower-ranked domains receive. Made-in-china.com and rsc.org sit among the highest AI-traffic earners on the cited list.

The distinguishing factor the study identifies is what a page offers after the answer ends. Engineerfix publishes broad engineering guides and FAQ-style explanations, content an AI system can summarize completely, leaving a reader with no unresolved reason to click. Made-in-china.com is a marketplace catalogue. Grainger is a distributor where a citation lands on a place to order, and one of its strongest AI traffic pages is a product listing for a Motorola two-way radio.

The Royal Society of Chemistry offers the sharpest split inside a single domain. Its most-cited pages sit in an /element/ subfolder, individual entries for neon, iron, lithium, chlorine and the rest of the periodic table. Its most-clicked pages sit in a /content/ subfolder holding peer-reviewed papers on PFAS chemistry, composite materials and biodegradable coatings. Citations land on the reference material. Clicks land on the research.

That distinction, between content an answer can absorb and content it can only point at, is the part of the study with the longest shelf life.

Why it matters for marketers

The manufacturing sector arrives at this measurement problem late, which makes its numbers a useful control case for everyone else.

The mention-citation gap Semrush recorded here is a sector-level instance of a pattern it documented across the wider economy. In its AI Visibility Index 2026, only 36 of more than 1,200 tracked brands held top-100 visibility on every platform in every month, and mention-source overlap ranged from 64% on Google AI Overviews to 30% on Gemini. Manufacturing sits toward the low end of that range, with two names out of fifteen on both lists.

Platform divergence has been measured repeatedly. A survey of 45 academic studies found URL-level overlap between organic Google results, AI Overviews and Gemini running between 0.11 and 0.18, and separate research put citation overlap between major platforms at 11% with roughly half of citation sets turning over monthly. The engineerfix case gives that abstraction a name and a traffic profile.

The traffic figure deserves the same caution. AI channels at 0.48% of manufacturing sessions is consistent with clickstream measurement elsewhere: Datos placed AI tools below 2% of desktop visits in the first quarter of 2026 while Google held 94% of search. Small referral share and large behavioural influence are not contradictory claims, and Similarweb's panel work found 55.9% of AI-influenced visits arriving through a later branded search rather than a click inside the conversation.

Against that, the click erosion on the organic side is not hypothetical. Ahrefs measured a 58% click-through reduction for top-ranking pages where AI Overviews appear, up from 34.5% a year earlier, and a randomized field experiment with 1,065 Chrome users produced a 39.8% drop in outbound organic clicks with a 34.5% rise in zero-click sessions. A sector whose generative coverage moved from 38% to 57% in seven months is exposed to that mechanic on an increasing share of its queries, including product and branded terms.

For industrial marketing teams, the practical reading of the dataset is that two budgets are now being evaluated against two different scoreboards. Brand investment shows up in mentions, where a hundred years of recognition still compounds. Content and distribution investment shows up in citations, where a directory with an Authority Score of 17 can outrank a global manufacturer. Neither scoreboard reports reliably into the analytics that finance departments read, which is the condition the study describes and does not solve.

Timeline

Summary

Who: Semrush, the search and AI visibility platform acquired by Adobe, with the study credited to Luke Harsel and contributor Nikola Manojlović. The brands measured include Boeing, 3M, John Deere, Vevor, Grainger, Siemens, Thermo Fisher, Carrier and the domains engineerfix.com, made-in-china.com and rsc.org.

What: A three-dataset study of AI-driven discovery in manufacturing and industrial search, finding AI Overviews on 57% of tracked search volume, AI channels at 0.48% of sessions, only two brands common to the top-15 mentions and top-15 citations lists, and no reliable correlation between citation volume and AI referral traffic.

When: The study was published September 8, 2026, with mention totals circulated to press on September 9. The underlying data covers January through July 2026.

Where: United States search behaviour, clickstream sessions across ten industrial categories, and AI answers on ChatGPT, Gemini, Google AI Mode and Google AI Overviews.

Why: Manufacturing marketing has been organised around ranking and click attribution for two decades. The data describes a sector where generative coverage of the results page is rising fast, referral traffic from AI remains under half a percent, the brands named in answers are not the sites being linked, and the influence of an AI conversation surfaces later as direct or branded traffic that conventional analytics record as unattributed.