AI halftime 2026: what actually moved and what’s still noise
Six months into 2026, AI moved real money: search traffic, market cap, token budgets, and headcounts. The problem is that almost none of it has been cleanly attributed yet. Kevin Indig’s H1 2026 halftime report, sent to 28,367 subscribers, maps what shifted and what’s still being misread.
The common thread across every major story: AI’s economic impact is expanding faster than anyone’s ability to measure it.
AI search: trust beats citations
Google’s AI Mode hit 1 billion monthly active users in H1 2026. Queries in AI Mode run roughly 3x longer than classic search queries. Google called it the biggest search box upgrade in 25 years at I/O 2026.
Indig’s research across the first half of the year surfaced five patterns operators and marketers should internalize:
- Measurement is broken. 91% of citations appear in only one of ChatGPT, Perplexity, or AI Overviews. Prompt tracking should be treated like polling and focus groups, not SEO rank tracking.
- Brand mentions outperform citations. Citations shape answers, but for most vendors and merchants, what matters is how often and in what context their brand appears across a panel of prompts.
- Trust is the top ranking factor. Close to 75% of consumers pick the number one result on an AI shortlist. But if they see a trusted brand anywhere on the list, they pick the trusted brand.
- AI Mode users accept recommendations at a high rate. 88% of the time, users accepted AI Mode product recommendations as the best available option. AI Overviews users, by contrast, click, evaluate, and compare, which matches classic search behavior.
- Technical and stylistic factors matter for agents. Leading with unique information, writing in a direct and easy-to-understand style, and keeping your site fast and accessible all improve AI visibility.
The practical reframe: AEO and GEO are brand channels, not performance channels. The unit of optimization is whether AI names, trusts, and recommends your brand, not whether you earn a citation link.
73.7 trillion tokens in 30 days
Starting in late 2025, Shopify, Uber, Meta, and other tech companies built internal token leaderboards that ranked engineers by how many AI tokens they consumed. The idea was to push adoption. What happened instead was waste at scale.
A Meta engineer created an internal leaderboard called Claudeonomics that ranked over 85,000 employees by token consumption. Employees left AI agents running on idle or useless tasks just to climb the ranks and earn titles like “Token Legend.” Meta engineers burned 73.7 trillion tokens in 30 days. The top user on the board averaged 281 billion tokens, which would cost over $1.4 million at standard API rates.
Jensen Huang put the intended logic plainly:
“If that $500,000 engineer did not consume at least $250,000 worth of tokens, I am going to be deeply alarmed.” — Jensen Huang
By April, CFOs had had enough. Most leaderboards were shut down after employees burned through their full-year token budgets in four months. The phase shifted from token maximizing to value maximizing.
What remains is an unmeasured step change in productivity. One side effect: the surge in Claude usage drove growth for AI infrastructure companies that Claude recommended, including Supabase.
SaaS fell 30%, but not because of performance
In February 2026, following the release of Opus 4.6 and Claude Cowork, the SaaS sector saw a market downturn that wiped out hundreds of billions in market cap over a few trading sessions. The SaaS vertical declined more than 30% on an annual basis, with revenue multiples compressing sharply from pandemic highs.
The important nuance, according to analysis cited by Indig: the decline is not correlated with actual company performance. It tracks with whether the market perceives a company’s product as vulnerable to AI disruption.
The bottom quartile of software companies is pulling down the entire index. The top quartile and the median IGV stock both outperformed the ETF over the same period. If you’re evaluating a role or investment in B2B software right now, market sentiment about AI robustness is the signal worth reading.
AI layoffs: the narrative doesn’t match the data
Challenger, Gray and Christmas reported that AI was the leading cited reason for job cuts in May 2026, with 87,714 cuts attributed to AI year-to-date through May, equal to 22% of all announced 2026 layoffs. Tech layoffs are up roughly 66% year over year, trending toward 150,000. Oracle cut 21,000 citing AI. Block cut 40% of its workforce and its stock rose, which Bloomberg framed as AI washing.
Indig’s read: AI is not the actual cause. He predicted in the H1 2025 report that AI layoffs were a PR narrative, and the follow-through data supports that. The real drivers are capital expenditure offsets, pandemic-era overhiring, and economic turbulence. Companies using AI as the stated reason for layoffs are, in a number of cases, rehiring those same employees.
The AI market fragmented fast
ChatGPT’s market share dropped from 78% in July 2025 to 56% in July 2026. Gemini grew from 15% to 30% over the same period. Claude went from 2% to 10%.
A few developments driving the shift:
- OpenAI pivoted toward enterprise, which now makes up roughly 40% of revenue and is heading toward 50% ahead of a potential IPO. OpenAI reportedly spends $2.8 billion per month to generate $1.1 billion in revenue, according to Ed Zitron.
- Anthropic’s conflict with the Pentagon pushed Claude to the number one free app on Apple’s U.S. App Store. Its longer-term position against open-source pressure is still being tested.
- SemiAnalysis found that a $200 plan from frontier model providers delivers the equivalent of $8,000 to $14,000 in tokens at standard API rates, which means the big labs are heavily subsidizing usage.
- Open-source models including Kimi K3, GLM 5.2, and Deepseek V4 are putting sustained pressure on U.S. labs, and their growth signals that the model layer is commoditizing while the application and harness layer is where value is building.
Forward-deployed engineer job postings rose 800% in the first nine months of 2025 and were still up 729% year over year in April 2026. Frontier labs are paying over $500,000 in total compensation for people who can turn a model demo into a working system inside a customer’s operation.
Publishers are taking it to court
Indig had predicted up to 70% of 2024 organic traffic could disappear by 2026. The actual number landed at roughly 33% referral loss year over year. 68% of Google searches now end without a click.
Publishers are responding through legal channels:
- A Munich court ruled Google liable for false statements generated by AI Overviews in June 2026.
- 400 newspapers sued OpenAI and Microsoft over unauthorized content use in June 2026.
- The UK CMA ordered Google to give publishers greater control and transparency over how their content is used in AI search, including opt-outs from AI Overviews, AI Mode, and Discover summaries, along with clearer attribution. Google responded by adding an impression-based AI report in Search Console.
USA Today CEO Mike Reed put the publisher calculus plainly: the company is approaching the point where blocking Google entirely and abandoning traditional search traffic becomes the more viable option.
The emerging replacement model is a content-for-training marketplace, with Parallel AI and Cloudflare both working to build infrastructure for it. The open questions: who sets the price, how do you prove a specific article contributed to an AI answer, and whether smaller publishers have any negotiation leverage at all.
Conde Nast CEO Roger Lynch offered the starkest planning posture: “Assume there’s no search. You have to have your businesses planned as if search is zero.”
The one-line summary
Every major H1 2026 AI story, whether search, tokens, SaaS valuations, layoffs, or publisher traffic, comes down to the same problem: impact is real but attribution is not. The operators who build measurement systems that can handle that ambiguity will have a structural advantage in H2.
I’m founder of BizStack at Cagri Sarigoz LLC and a passionate advocate for entrepreneurs.
With over 14 years in tech, marketing, and AI, including my role as Head of SEO at CitizenShipper and co-founder of TaleBot at Intale AI, I’m dedicated to sharing genuine, useful product insights and tips.
At BizStack, I aim to cut through the digital noise to provide clear, actionable advice.
And more than all else, I’m a father to a (always) little girl and a husband.
Contact me at [email protected] for assistance.
