As 2025 draws to a close, it's time to review the ambitious AI predictions made at the start of the year. This annual audit assesses where artificial intelligence truly transformed work, search, and strategy, and where its anticipated impact fell short. Out of ten forecasts, the year saw three misses, two mixed results, and five significant hits, offering crucial insights into the rapid evolution of AI deployment across various industries.
Here's a detailed breakdown of the 2025 predictions and their outcomes:
1. Agentic LLM Models Reach +100 Million Users
Score: Miss
Thought Process: The initial assumption was that as large language models (LLMs) improved in reasoning, user behavior would shift from simple chat interactions to autonomous actions, making agentic AI the logical next step for mass adoption.
Reality: While general LLM usage, exemplified by platforms like Gemini and ChatGPT, soared past 800 million weekly users, the adoption of truly autonomous agents remained a niche feature. These agents, designed to perform complex actions such as "buying a product" without direct oversight, struggled with "wonky" performance and were primarily used by power users. Google's "Project Mariner" and OpenAI's agent features only entered broad public beta in mid-2025, indicating a slower-than-expected productization and a general consumer preference for AI as an information tool rather than an autonomous actor. User trust in letting software manage finances or act unsupervised proved a significant hurdle.
2. More AI Victims
Score: Hit
Thought Process: This prediction stemmed from early signals observed in industries like online education (Chegg) and developer forums (Stack Overflow), viewed as initial signs of a broader margin collapse. The hypothesis was that any labor-intensive industry with AI positioned between buyers and services would experience significant disruption.
Reality: By Q3 2025, major call center outsourcing firms faced a severe crisis. Enterprise clients aggressively transitioned to "AI Voice First" support layers, drastically reducing the demand for human agents. Similarly, the translation services sector continued to shrink as browser-based, real-time AI translation became a native feature in operating system updates. RWS Holdings, a prominent translation service provider, reported a 41% plummet in Adjusted EBITDA and nearly a 60% drop in profit before tax in its 2025 half-year report. Chegg continued its decline, and Concentrix shares dropped approximately 9% in a single day in September 2025 after missing earnings expectations, as the traditional "per-seat" billing model for Business Process Outsourcing (BPO) was dismantled by AI-driven solutions.
3. AI Automation Becomes The Default For Marketing Teams
Score: Hit
Thought Process: This forecast was based on observing clients integrating tools like AirOps, Make, Zapier, and custom scripts to maintain or shrink headcount. The expectation was that economic pressures combined with improved tooling would push marketing teams towards "systems thinking," prioritizing workflow automation over channel-specific strategies.
Reality: "System building" emerged as a primary skill in marketing job descriptions throughout 2025. Recruitment data from Ashdown Group's 2025 Marketing Job Market Report indicated that roles requiring campaign automation and AI tool integration commanded a 7-9% salary premium. A 2025 HubSpot State of Marketing report further revealed that 78% of B2B organizations now rely on marketing automation as their core infrastructure. With marketing budgets remaining tight, the "Team of One" structure, powered by automation chains and custom AI workflows, became the industry standard.
4. AI Overviews Evolve
Score: Hit
Thought Process: AI Overviews were seen as an experimental feature, not a final product. The belief was that Google would iterate towards more personalized and richer Search Engine Results Page (SERP) formats as multimodal AI models matured, primarily to defend its market position against standalone LLMs.
Reality: Google extensively tested its "Web Guide" SERP layout as of November 2025. For users opted into the Search Generative Experience (SGE), this became the default. Instead of traditional blue links or a single AI answer, the "Web Guide" organizes search results into AI-generated "buckets" or headlines. Key developments included Google embedding YouTube Shorts and timestamped video clips directly into AI Overviews in June 2025, and deeper integration of "AI Mode" into AI Overviews by December.
5. Reddit Becomes Part Of The Default Channel Mix
Score: Hit
Thought Process: The prediction was that as Reddit's visibility grew across product searches and troubleshooting queries, marketers would be compelled to treat it as a core performance channel rather than a secondary project. Its high visibility in search results compared to ad revenue potential on other platforms suggested significant untapped upside.
Reality: Reddit experienced a banner year in 2025. Ad revenue surged by 61% year-over-year in Q1 2025, followed by another 68% increase in Q3, reaching $585 million. With Google Search increasingly prioritizing forum discussions, Reddit became an indispensable placement for advertisers seeking high-intent traffic. Daily Active Users (DAUs) reached 108.1 million in Q1 and climbed to 116 million by Q3 2025. Reddit is now the second-largest site on Google by visibility, surpassed only by Wikipedia. The launch of true e-commerce catalog ads (Dynamic Product Ads – DPA) in Q1 2025 beta tests showed a 2x higher return on ad spend (ROAS) compared to previous formats, making Reddit viable for performance marketers, not just brand awareness teams. Furthermore, Reddit remains the most cited platform for most LLMs.
6. More Sites Cloak For LLMs
Score: Miss
Thought Process: The expectation was that B2B sites, in particular, would quickly and tactically adapt by feeding bots cleaner, more structured versions of their content, assuming LLMs would reward such patterns and that the benefits would outweigh any perceived bending of rules.
Reality: The envisioned "Bot-Only Web" did not emerge through cloaking. Instead, it manifested via APIs and paywalls. Most major publishers aggressively blocked bots using robots.txt files and pursued legal action, as seen in the ongoing dispute between The New York Times and OpenAI. This indicates a preference for control and monetization through structured access rather than covert optimization.
7. The Current Google Shopping Tab Will Become The Default
Score: Miss
Thought Process: The Shopping tab was viewed as Google's experimental ground for a future Amazon competitor, similar to how past tab experiments eventually integrated into the main SERP. The core belief was that Google would intensify its push for shoppable, personalized results in response to AI competition and the growing demands of e-commerce.
Reality: Google maintained a distinct separation for its main search tab. While the Shopping experience did become more personalized and AI-driven, often resembling a feed, Google did not replace the default search experience with the Shopping tab interface for commercial queries. The two functionalities remained largely separate.
8. AI-Generated Audio And Video Hits Mass Adoption
Score: Hit
Thought Process: This prediction connected the rapid advancements in generative AI tools with the constant pressure on creators to produce more content with limited budgets. It was assumed that once tools like Sora, Veo, and ElevenLabs achieved a basic quality threshold, they would inevitably integrate into production pipelines, even if audiences could still discern synthetic elements.
Reality: 2025 was indeed the year of the "Synthetic Creator." YouTube was compelled to update its Partner Program (YPP) policies in July 2025 specifically to manage the influx of AI-generated content. While the policy cracked down on "Mass-Produced & Repetitious Synthetic Content" (often termed "AI slop"), it explicitly protected creators who used AI tools for "production assistance" – such as B-roll, voiceovers, and scripting – provided there was "clear editorial oversight." This distinction allowed high-quality AI-generated elements to become standard for millions of creators. Reports indicate that 87% of creators now incorporate AI into their workflows.
9. Google And Apple Divorce
Score: Mixed
Thought Process: The Department of Justice (DOJ) case against Google was perceived as a structural threat to Google's distribution deals, with the expectation that judges would eventually challenge the default search arrangements. The framing was that even a partial unwinding of exclusivity could shift how search power is negotiated, without instantly determining a "winner" in the search market. Judge Mehta's initial conclusions suggested a firm stance on remedies.
Reality: The DOJ remedy ruling in September 2025 proved largely disappointing, described as a "toothless tiger." While the court prohibited Google from paying for default exclusivity on browsers (Chrome/Safari) and devices, key loopholes remained. Google can still pay Apple to be the default, provided the contract does not explicitly state "Apple must not use anyone else." Crucially, the judge did not enforce a "choice screen" similar to those in the European Union, leaving Apple free to voluntarily implement choice screens or offer alternatives (like ChatGPT or Perplexity) without fully sacrificing Google's payments. Furthermore, Apple is now licensing Gemini for Siri, indicating a continued, albeit modified, partnership. The outcome was less a "divorce" and more a forced transition to an "open marriage."
10. Apple Or OpenAI Announces Smart Glasses
Score: Mixed
Thought Process: This prediction viewed smart glasses as the logical next hardware interface for AI assistants, especially given Meta's traction in the AR/VR space. It was assumed that either an OpenAI-Jony Ive collaboration or Apple's need for a new device narrative would bring a prototype into public view, even if mass adoption was years away.
Reality: On November 24, 2025, OpenAI CEO Sam Altman and designer Jony Ive officially confirmed their joint hardware venture (under the startup "LoveFrom") had a finished prototype during an interview hosted by Laurene Powell Jobs. The device was described as "screen-free" and "less intrusive than a phone," aligning with the "smart glasses" or "AI Pin" form factor. However, it's not yet definitively proven that OpenAI's product will be glasses; it could be a necklace-style device. Meanwhile, Apple did not announce a new product, grappling instead with internal leadership issues. Adding another twist, just before this memo's publication, Google announced new smart glasses slated for 2026.
My Conclusion: 2025 Was The Year AI Deployment Began
For professionals in tech and digital marketing, 2025 will be remembered as the year AI-driven "pilot programs" concluded, and official "deployment" truly began. This shift wasn't confined to internal team workflows and tech stacks; it also fundamentally transformed classic search habits, which had been shaped by decades of human-search engine interaction.
While we didn't quite reach the sci-fi future of autonomous agents managing our groceries (Prediction No. 1) or widespread smart glasses (Prediction No. 10), we witnessed something more pragmatically disruptive: a world where marketing teams are leaner but more technically proficient, where Business Process Outsourcing (BPO) industries are collapsing under AI pressure, and where "Googling it" increasingly means "Reading a Reddit thread summarized by an AI."
Featured Image: Paulo Bobita/Search Engine Journal
