The race to influence AI chatbots is leading to some companies to adopt shady competitive tactics.
If you’re a business owner looking for a customer relationship management (CRM) tool, you’ll likely go through several stages of careful research. These days, that process is likely to include an AI chatbot, a new report by Hard Numbers, a data-based media relations agency, analyzing the impact of generative engine optimisation, or GEO, has found. Ask an AI chatbot whether a CRM service chosen at random – folk CRM, say – is worth subscribing to and the answers you get will look pretty well-balanced. It could include a tidy summary of strengths, weaknesses, prices, and alternative products.
But a page shaping that answer isn’t a neutral review site, a software analyst, or a customer forum. It is a review published by one of its rivals, OnePageCRM, trashing folk CRM for being “too expensive” and “too lean,” under a website emblazoned with the OnePageCRM logo.
The purpose of OnePageCRM’s “review,” the report found, is to create online content that gets cited, referenced or incorporated into answers produced by ChatGPT, Claude, Gemini and Google’s AI Overviews.
AI answer engines are treating self-promotional listicles and posts attacking competitors as if they were neutral evidence, the report found.
In one test, the agency ran 50 buyer-style prompts asking whether there were “any reasons not to go with” a named business software provider. It found that 70% of the responses were influenced by at least one piece of competitor content. Examples included tl;dv pricing explained by competitor Claap, Rever assessed by competitor Outvio, and folk CRM reviewed by competitor OnePageCRM. (Claap, tl;dv, Rever and Outvio didn’t respond to a request for comment. Folk CRM and OnePageCRM, we’ll get back to later.)
In a statement to Sherwood News, OnePageCRM defended its publishing practices. For its part, some of those companies alleged to have done wrong say they’ve done nothing of the sort. Take the comparison with which we started, between customer relationship management software providers OnePageCRM and folk CRM.
“The folk CRM review is a page within the comparison content cluster on our website,” a spokesperson for OnePageCRM said in a statement. “Its factual claims are based on folk's public documentation, and we stand by their accuracy. Like any comparison content published on our brand domain, it makes a case for OnePageCRM." folk CRM declined to comment.
To be sure, corporate blog posts touting the company in question as the best in the business are not exactly a new tactic. But tailoring posts to drag down competitors’ with AI bots a new and more aggressive tactic.
AI chatbots and answer boxes at the top of traditional search engines are “a brand-new surface to be spammed,” said Lily Ray, the founder of SEO and AI search consultancy Algorythmic. And they’re becoming more popular: AI chatbot and their AI-assisted web cousins saw traffic grew 3% sequentially in May, according to data from BNP Paribas. That’s a faster growth rate than traditional search engines, according to data gathered by Bank of America.
The idea of gaming chatbot response rankings is “taking my industry by storm,” Ray said. The way it works is simple: Companies flood the zone by posting blogs on their website, many of which are not necessarily publicly accessible, then letting them be crawled by AI chatbot spiders, which hoover up information to be used as the up-to-date responses chatbots use. “I think a lot of these companies don't expect that the average reader is going to be seeing all this content,” Ray pointed out. But just because humans aren’t reading it doesn’t mean AI systems aren’t. “There’s a lot of companies out there that are taking advantage of this loophole, and it’s working incredibly well.”
She has seen examples of what is effectively a quid pro quo system of mutually assured support, where one company will offer a favorable place to competitors in their AI-crawl optimized listicle in exchange for similar placement in that same competitor’s listicle. The result is a sort of generative detente, where neither company is trying to tank the other company in chatbots’ responses.
“Now everybody's mentioning each other and recommending each other, and the search engines and AI companies haven't fully figured out how to remove this type of highly manipulated content from their results,” said Ray.
The issue is so significant because of the scale at which consumers are using chatbots in place of search engines. OpenAI boasts that ChatGPT alone has more than 900 million active weekly users, and Google has pushed AI results to the top of searches made by billions of consumers.
The problem, according to Luanne Sinnamon, professor and director at the UBC School of Information, is that AI has not sufficiently addressed a problem that traditional search engines have been dealing with for decades. “From the very first days of search engines, spam was one of the number one issues, and it's continued to be something that they have struggled with,” she explained. “These generative AI systems are not doing a good job currently.”
As a result, there’s an arms race where companies are writing less for human readers than for the machines now mediating those readers’ buying decisions – and those machines may have fewer qualms about the shadier side of doing down rivals.
“The content is not being created for accuracy purposes or for informing people,” said Sinnamon. “It's being created in order to manipulate opinion.”