By Amanda Leon, founder of OilMoney.io. Houston-raised, came up in the Houston marketing team of Vallourec, and has run Unrth, a brand and growth studio for oil and gas and industrial companies, since 2017, with her boots on client locations in Midland and Odessa. Last reviewed October 30, 2026.
The B2B purchase process used to start with a search page, a trade show, or a phone call. More and more, it starts with an AI answer, according to Matthew Bertram, whose team studies how AI tools describe oil and gas companies. A buyer can get a short list before any rep knows there's a project. The field voices argued about that shift all year. The research on how buyers search, and whom they trust, shows what still decides the deal.
What the Field Voices Said
Matthew Bertram made the case most often. In March, Bertram wrote that buyers, investors, recruits, and partners now learn about oil and gas companies through AI answers, not only Google rankings, and that AI tools select trusted sources rather than ranking pages, so a service company can lead Google for years and still be missing from AI answers. In October, Bertram opened a talk at a Goldman Sachs 10,000 Small Businesses event with the point that AI tools are now the biggest influence on whether a customer chooses a business.
Bertram also brought the most useful number. The team ran 12,778 answers across five AI engines over 164 days, and ChatGPT returned the same set of citations twice for an identical prompt just 0.06% of the time. Bertram's conclusion: a single AI visibility report is a roll of the dice.
Collin McLelland, CEO of Collide, told oil and gas CEOs not to hire AI engineers, because they run oil companies, not AI companies. Chuck Yates described what happens on the ground when AI arrives: the CEO loves it, and the engineer in the third row decides it's a threat. And on the Oil and Gas Sales and Marketing Podcast, the hosts put the share of research buyers do before talking to a seller at about 70% to 75%, while saying they didn't recall the exact figure.
How a Cognitive Scientist Reads an AI Answer
I studied cognitive science, with an emphasis in neuroscience, and diplomacy and world affairs at Occidental College from 2009 to 2013. Cognitive science draws on psychology, neuroscience, and computer science, so how a machine picks an answer and how a person picks a vendor are questions the field asks side by side. My liberal arts training taught me to bring the principles that work in one field into another. AI search is a good place to do it, because the machine and the buyer turn out to look for some of the same things.
Buyers Search More When the Purchase Is New
We looked for a peer-reviewed source for the 70% figure and didn't find one, and the hosts flagged their own uncertainty. What the research does show is narrower, and more useful. Lüders and colleagues interviewed 40 purchasing managers and found that how much buyers search online depends on the purchase. Online search mattered most when the purchase was new to them. Their own experience shaped how they used what they found.
In oilfield terms: for a routine reorder, the buyer calls who they know. For a new problem, a new play, or a new kind of tool, they search first, and that's where AI answers come in.
Searching more didn't make buyers less loyal to reputation. When Krijestorac, Garg, and Konana surveyed 196 purchasing managers, the ones who leaned hardest on digital sources were more open to new products, and they also relied more on shortcuts like brand and peer opinion. They felt more attached to their current vendors, too, yet were less likely to buy from them again. More research online made reputation matter more, and made loyalty easier to lose.
The Top Answer Gets the Trust
In an eye-tracking study, Pan and colleagues found that searchers trusted Google's ranking so much that they clicked the higher results even when the summaries below them were more relevant. An AI answer is that top position with no page two.
The older research on credibility explains what earns the spot. In Hovland and Weiss's 1951 experiments, the same message changed more minds when it came from a credible source. A few weeks later, that gap had mostly closed, which is a good argument for staying visible. Pornpitakpan's review of five decades of these studies found credible sources usually win, with conditions where credibility matters less. Bertram's point fits right on top of this research: AI tools are built to find credible sources, and buyers trust what the tools find.
What Gets a Company Into an AI Answer
Computer scientists have started testing this. In a peer-reviewed paper at the 2024 KDD conference, Aggarwal and colleagues rewrote web content in different ways and measured how often generative engines surfaced it. Some changes raised a source's visibility in AI answers by up to 40%, and what worked varied by subject. The strongest changes made content easier to cite: adding sources, quotations, and statistics. For a rep, that means specifics. The job, the date, the result, and the number beat a page of adjectives.
Bertram's 0.06% has support too. Ouyang and colleagues gave ChatGPT the same 829 coding problems repeatedly and found it often returned different answers to the same request. On one benchmark, 75.76% of tasks never produced the same test output twice, and even the setting meant to make answers repeatable didn't guarantee it. That's code, not vendor lists, but the lesson carries: one AI answer is a sample, not a verdict.
Why Engineers Distrust the Machine
Dietvorst, Simmons, and Massey found people lose confidence in an algorithm faster than in a person after watching both make the same mistake, often enough to pick a worse human forecaster. Logg, Minson, and Moore found the flip side: lay people often followed advice more when they thought it came from an algorithm. Experts relied on it less, even when that hurt their accuracy.
That's the engineer in Yates's third row. Engineers are experts, and many have watched a tool fail on location. So the AI tool may build a buyer's short list, and the engineer will still check it against people they trust. Our answers on AI vendor shortlists and whether AI will replace oil and gas sales go further.
What AI Can't Replace
The one thing AI can't replace is the lived-in story, the real “come with us on this journey.” I studied neuromarketing, and the best way I know to understand people is to watch what they actually do, which is ethnography, not a survey. AI can summarize what's been written about you. It can't stand on location with your customer.
For more than seven years I've helped criminal defense lawyers in California win clients. In consumer services generally, people do their research alone before they pick up the phone, and the call goes to the name that felt trustworthy. Oilfield buyers now have a research assistant that never sleeps. The call still goes to a person.
Where the Science Has Limits
The research on AI search is young. Peer-reviewed work on how oil and gas buyers use AI tools is thin. The purchasing studies here cover online search in general.
The visibility study used a test set of queries. Up to 40% is a ceiling, not a promise, and results varied by subject. Nobody can guarantee a spot in an AI answer.
The repeatability study was about code. It backs the idea that AI answers vary. It didn't measure vendor recommendations.
Pan's searchers were college students. Experienced buyers may scrutinize more, as the algorithm studies suggest.
Credibility fades without reminders. In Hovland and Weiss's data, the credible source's edge shrank within weeks. The related sleeper effect, where a weak source's message catches up later, has been hard to replicate, so I lean only on the fading.
The 70% figure is unsourced. The direction fits the research. The number is the hosts' recollection.
How to Show Up When AI Answers First
Be worth citing. Post specifics: the job, the date, the result, the number. Vague praise gives an AI tool nothing to quote.
Don't trust one AI report. Ask the same question many times, the way Bertram's team did, before you believe what it says about you.
Get named by other people. Buyers who research online lean on brand and peer opinion, so customers and peers who mention you count twice.
Keep the field story human. A result on location, told by the person who was there, is the part no tool can write.
Respect the engineer's skepticism. Bring proof that holds up when an expert checks it.
Be known before the search. If a buyer has heard your name before the AI tool says it, the answer confirms what they already think. That's the idea behind The Warm Walk-In™. Start with your LinkedIn profile, and read our take on AI LinkedIn outreach before you automate anything.
Follow Along
Want the short version in your inbox? The Tailgate is one letter a month from the patch. Sign up at the bottom of any page. If you'd rather see what a buyer sees when they look you up, get your Bison Walk-Around.
Sources and As-Of Dates
Peer-reviewed studies (each DOI checked on Crossref, October 30, 2026):
Lüders, M., Klarmann, M., Wouters, M., and Gerlach, A. (2023). How online information search behavior and the role of tacit knowledge differ across clusters of purchase situations. Journal of Purchasing and Supply Management, 29(4), 100862. Supports: online search matters most for purchases that are new to the buyer (40 purchasing managers).
Krijestorac, H., Garg, R., and Konana, P. (2021). Decisions Under the Illusion of Objectivity: Digital Embeddedness and B2B Purchasing. Production and Operations Management, 30(7), 2232-2251. Supports: digitally reliant buyers lean more on brand and peer opinion (196 purchasing managers).
Pan, B., Hembrooke, H., Joachims, T., Lorigo, L., Gay, G., and Granka, L. (2007). In Google We Trust: Users’ Decisions on Rank, Position, and Relevance. Journal of Computer-Mediated Communication, 12(3), 801-823. Supports: searchers trust top-ranked results even when lower ones are more relevant.
Hovland, C. I., and Weiss, W. (1951). The Influence of Source Credibility on Communication Effectiveness. Public Opinion Quarterly, 15(4), 635-650. Supports: the same message persuades more from a credible source, and the gap fades over weeks.
Pornpitakpan, C. (2004). The Persuasiveness of Source Credibility: A Critical Review of Five Decades' Evidence. Journal of Applied Social Psychology, 34(2), 243-281. Supports: five decades of evidence that credible sources usually persuade more.
Aggarwal, P., Murahari, V., Rajpurohit, T., Kalyan, A., Narasimhan, K., and Deshpande, A. (2024). GEO: Generative Engine Optimization. Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 5-16. Supports: content changes raised visibility in generative engine answers by up to 40%.
Ouyang, S., Zhang, J. M., Harman, M., and Wang, M. (2025). An Empirical Study of the Non-Determinism of ChatGPT in Code Generation. ACM Transactions on Software Engineering and Methodology, 34(2), 1-28. Supports: ChatGPT gives different outputs to the same request.
Dietvorst, B. J., Simmons, J. P., and Massey, C. (2015). Algorithm aversion: People erroneously avoid algorithms after seeing them err. Journal of Experimental Psychology: General, 144(1), 114-126. Supports: people drop algorithms after seeing them err.
Logg, J. M., Minson, J. A., and Moore, D. A. (2019). Algorithm appreciation: People prefer algorithmic to human judgment. Organizational Behavior and Human Decision Processes, 151, 90-103. Supports: lay people favor algorithm advice; experts rely on it less.
Field voices:
Matthew Bertram, LinkedIn posts of March 18, March 30, September 18, and October 21, 2026.
Collin McLelland, LinkedIn post of April 13, 2026.
Chuck Yates, LinkedIn post of August 4, 2026.
Oil and Gas Sales and Marketing Podcast, Mastering Buyer Champions (July 29, 2026) and AI and Data Governance: The Future of Sales and Marketing (July 8, 2026), for the 70% to 75% figure, which the hosts gave without a source.
The Warm Walk-In™ is an original method created by Amanda Leon and owned by The Valours Creative Innovation LLC.
