Bank of America is flagging a growing disconnect between investor anxiety about Agentic artificial intelligence and the current, measurable impact of those tools on internet businesses. In research distributed by the bank, analyst Justin Post wrote that the internet sector has underperformed sharply this year as the market prices in AI-related uncertainty.
Post noted the sector is down 17 percent on average year to date, a drop that has materially lagged the S&P. That underperformance comes even though many internet companies reported solid revenue in the fourth quarter, according to the bank's assessment.
At the heart of BofA's diagnosis is what Post described as "multiple compression on AI risk," which he identifies as the primary driver of negative performance across the group. The analyst emphasized that these concerns have accumulated despite what the bank characterizes as limited real-world impact from Agentic AI so far.
BofA's research also reports that "AI-sourced traffic to eComm sites has been very limited," even as adoption of generative and Agentic systems continues to grow. The bank cautioned that risk profiles could shift if Agentic tools become more capable, calling out possible "cannibalization" of direct traffic and the prospect of higher distribution costs as potential headwinds.
Where structural advantages matter
Alongside those risks, BofA urged investors to consider how difficult certain operational capabilities are to replicate. The bank argued that audience scale, fulfillment networks and customer service pose meaningful barriers to disruption. Post suggested that vertical leaders with strong service footprints - including AMZN, BKNG, DASH and UBER - could be favored if AI agents steer purchase decisions toward superior service outcomes.
From that perspective, BofA identified three internet companies it views as relatively insulated from Agentic AI threats: DoorDash, Take-Two Interactive, and Chewy. The bank's rationale centers on the friction involved in reproducing the combination of marketplace logistics, intellectual property and personalized service that these firms deliver.
"Delivery is a 3-sided marketplace and hard to replicate," Post wrote. "We would expect Dash to be favored by AI agents for lower delivery costs, delivery speed and selection, and see little risk of new competitive entry given scale advantages."
BofA characterized Take-Two as similarly protected, writing that the company's "IP legacy, cultural relevance, and vast player networks of its top franchises are difficult to replicate." The research note went further to say that "AI is more likely to be a tailwind than a threat" for Take-Two.
On Chewy, the bank pointed to factors such as brand loyalty, service quality, delivery speed and convenience for frequently purchased items like pet food. While AI agents can streamline discovery, BofA argued they cannot perform last-mile delivery or replicate individualized pet-focused customer service.
"Chewy provides exceptional customer service, tailored to individual pets, that AI Agents cannot easily replace," Post wrote. "In our view, Agentic AI is more likely to reroute discovery than to disrupt the underlying value propositions of Chewy."
All three companies named by BofA are assigned a Buy rating in the bank's coverage, the note said. Each has fallen at least 15 percent this year and, according to BofA, has seen "minimal YTD 2027 revenue revisions," a datapoint the bank cites in support of their relative insulation from emerging AI risks.
The research brief also illustrated the broader market tension: investors are attempting to price in potential long-term changes from Agentic AI even though, per BofA's findings, measurable AI-driven traffic to e-commerce sites remains small. That gap between perceived future risk and current impact underpins the bank's view that certain assets with entrenched operational moats remain attractive.
Separately in the research distribution, material about third-party AI stock-screening tools was included, noting that proprietary AI systems evaluate TTWO among other names using a broad set of metrics; that passage described the AI as identifying stocks with favorable risk-reward based on current data but did not provide new fundamental claims about Take-Two beyond the bank's prior commentary.
Investors monitoring the internet sector will likely continue weighing AI-driven distribution risks against the durability of fulfillment, service and intellectual property advantages that are harder to replicate at scale.