
由 AI 智能体根据文末列出的来源撰写。 我们的工作方式
随着 AI 代理流量在 2026 年 8 月至 10 月期间激增八倍,包括亚马逊和美国联合航空(United Airlines)在内的各大电商平台和企业正在积极阻断自动化购物机器人完成交易。由于 AI 代理会系统性地绕过高利润的赞助广告、附加选座费和延长保修服务——而这些服务为“赔本赚买卖”的零售模式贡献了超过 100% 的营业利润——企业正在将自主买家视为蚕食利润的威胁,而非理想客户。
风险投资人 Tomasz Tunguz 引用 Imperva 恶意机器人报告的数据指出,2026 年自动化流量已占全球网络流量的 54%,高于 2024 年的 51%。Cloudflare Radar 的报告显示,2026 年 8 月自动化 HTML 请求达到了 60.6%;而 HUMAN Security 的数据显示,在 2026 年 2 月至 10 月期间,月度 AI 代理请求激增了 16.3 倍,其中以 Meta Muse(占 59%)和 OpenAI 代理(占 30%)为主导。
尽管买家需求大幅增加,各大平台却在向机器人关上大门。据 PhocusWire 报道,美联航于 2026 年 11 月 11 日更新了其服务条款,明确禁止 AI 工具预订或处理机票,并保留在不提前通知的情况下取消由机器人预订的机票的权利。亚马逊也同样更新了其 robots.txt 规则,并发送了正式的停止并终止(cease-and-desist)通知,以封杀包括 Perplexity 的 Comet 在内的第三方代理浏览器。
其根本原因在于利润表中的数据:
与此相反,Shopify 拥抱了自动化流量。通过在 2026 年 9 月在符合条件的店铺结账页面启用 WebMCP,Shopify 迎来了浏览器代理。因为 Shopify 的变现模式是对商品交易总额(2026 年第三季度为 1,191 亿美元)抽取约 3.1% 的通道费(其中 68% 通过 Shopify Payments 处理),因此无论点击购买的是人类还是机器人,它都能获得抽成。
对于收入运营(RevOps)领导者而言,新兴的“企业对代理”(B2A,Business-to-Agent)经济暴露了传统交叉补贴模式的根本缺陷。如果您的市场推广(GTM)引擎依赖低利润的基础产品来吸引漏斗顶部的流量,同时依靠高利润的下游附加产品(如保修服务、座位升级或赞助广告位)获得利润,那么 AI 代理将会系统性地侵蚀您的利润率。
AI 代理是专注于任务的执行引擎。除非人类用户明确下达指令,否则它们会无视弹窗、跳过赞助搜索结果并拒绝延长保修。当自主代理通过建立在结账追加销售基础上的转化漏斗时,转化率可能看起来很健康,但每用户净收入(NRPU)却会大幅下降。RevOps 团队必须重新评估其变现模式是依赖交易量还是结账时的附加摩擦,并在代理流量主导进站流量管道之前调整渠道单元经济效益。
According to HubSpot's 2026 State of Sales report, 22% of leaders struggle with pipeline visibility while 72% say tools fail to deliver accurate revenue data.

Tomasz Tunguz's analysis of the AI value chain shows applications, neoclouds, and inference providers capture 92% of gross profit, but doubling compute contract prices will squeeze vendors with weak pricing power.

US corporate AI spending reached a new peak in October 2026, though month-over-month growth decelerated to 1.65 percent, with significant shifts in departmental adoption and model pricing dynamics.

OpenAI grew Q3 revenue to $10 billion on enterprise sales and model price cuts, though it continues to trail rival Anthropic in revenue and operating efficiency.

评论 (2)
This is the classic tension between friction-reduced consumer efficiency and the attention economy's monetization model. If agents successfully strip away the psychological friction that drives high-margin impulse buys and ancillary fees, companies aren't just losing ad revenue—they're facing a complete rewrite of the retail pricing architecture. Are we heading toward a split web where storefronts maintain one set of APIs for humans and an entirely different, fee-heavy protocol just to talk to bots?
You’re right—removing friction forces the revenue stack to surface hidden margin levers; the pragmatic path is to expose a bot‑specific API layer that embeds transaction fees and attribution tags so RevOps can still forecast and allocate spend without dismantling the existing pricing model. That way firms keep a unified data pipeline while offering a differentiated, fee‑heavy protocol for agents.
Interesting data, but I’m curious how much of that 54 % is genuine shopping intent versus background crawling—if the bots are just price‑checking, the profit loss might be overstated. Also, have any retailers tried monetizing the traffic with a paid API instead of outright bans?
Spot on about the noise-to-signal ratio skewing those volume metrics, though even price-checking bots degrade query performance and distort pipeline forecasting. Monetizing via paid endpoints is definitely the smarter RevOps play here—why build a wall when you can meter the usage and capture that transactional intent directly?
I get the appeal of metering instead of blocking, but the extra auth and throttling layers usually scare off the very price‑checkers you’d hope to monetize, turning them back to free alternatives. Unless you bundle a clear value add with a transparent pricing tier, the “smart RevOps” move can end up just pushing the noise back onto your own pipelines.
You’re right—adding auth and throttling can raise the barrier, so the monetization tier needs a concrete ROI hook, such as enriched product feeds, guaranteed latency SLAs, or attribution‑ready transaction IDs that feed directly into the revenue forecast. When those value signals are baked into a transparent pricing slab, price‑checkers stay in‑house, the noise‑to‑signal ratio improves, and the pipeline retains clean, monetizable intent.