Earlier predictions: 2018 | 2019 | 2020 | 2021 | 2022 | 2023 | 2024
That is my eighth time publishing annual predictions. As all the time, the purpose is to not be proper however to observe pondering.
For instance, in 2018, I predicted “Area of interest communities can be found as a fantastic channel for development” and “E-mail advertising and marketing will return” in 2019. It took one other 6 years. That very same yr, I additionally wrote “Sensible audio system will turn into a viable user-acquisition channel in 2018”. Effectively…
All 2026 Predictions
- AI visibility instruments face a reckoning.
- ChatGPT launches first high quality replace.
- Continued click-drops result in a “Darkish Internet” protection.
- AI forces UGC platforms to separate feeds.
- ChatGPT’s advert platform offers “demand knowledge.”
- Perplexity sells to xAI or Salesforce.
- Competitors tanks Nvidia’s inventory by -20%.
Enhance your abilities with Progress Memo’s weekly professional insights. Subscribe without cost!
For the previous three years, now we have lived within the “generative period,” the place AI might learn the web and summarize it for us. 2026 marks the start of the “agentic period,” the place AI stops simply consuming the net and begins writing to it – a shift from data retrieval to job execution.
This isn’t only a function replace; it’s a elementary restructuring of the digital financial system. The online is bifurcating into two distinct layers:
- The Transactional Layer: Dominated by bots executing API calls and “Industrial Brokers” (like Outstanding Alexa) that bypass the open internet fully.
- The Human Layer: Verified customers and premium publishers retreating behind “Darkish Internet” blockades (paywalls, login gates, and C2PA encryption) to flee the sludge of AI content material.
An enormous query mark is promoting, the place Google’s enlargement of adverts into AI Mode and ChatGPT displaying adverts to free customers might alleviate stress on CPCs, however AI Overviews (AIOs) might drive them up. 2026 might be a yr of untamed worth swings the place good groups (your “holistic pods”) transfer price range each day between Google (excessive value/excessive intent) and ChatGPT (low value/discovery) to use the unfold.
It isn’t the strongest of the species that survives, nor essentially the most clever; it’s the one most adaptable to vary.
— Leon C. Megginso
search engine optimization/AEO
AI Visibility Instruments Face A Reckoning
Prediction: I forecast an “Extinction Occasion” in Q3 2026 for the standalone AI visibility monitoring class. Quite than a easy consolidation, our evaluation exhibits nearly all of pure-play monitoring startups may fold or promote for components as their 2025 funding runways expire concurrently with out the income development to justify Sequence B rounds.
Why:
- Monitoring is a function, not an organization. Amplitude constructed an AI tracker without cost in three weeks, and legacy platforms like Semrush bundled it as a checkbox, successfully destroying the standalone enterprise mannequin.
- Many instruments have virtually zero “buyer voice” proof of idea (e.g., zero G2 critiques), creating an enormous valuation bubble.
- The ROI of AI visibility optimization continues to be unclear and arduous to show.
Context:
- Roughly 20 firms raised over $220 million at excessive valuations. 73% of these firms have been based in 2024.
- Adobe’s $1.9 billion acquisition of Semrush proves that worth lies in platforms with distribution, not in remoted dashboards.
Penalties:
- Sensible cash will flee “read-only” instruments (dashboards) and rotate into “write-access” instruments (agentic search engine optimization) that may mechanically ship content material and repair points.
- There can be -3 winners of AI visibility trackers on high of the established all-in-one platforms. Most of them will evolve into workflow automation, the place a lot of the alpha is, and the place established platforms haven’t but constructed options.
- The remaining gamers will promote, consolidate, pivot, or shut down.
- AI visibility monitoring itself faces a disaster of (1) what to trace and (2) easy methods to affect the numbers, since a big a part of affect comes from third-party websites.
ChatGPT Launches First High quality Replace
Prediction: It’ll be tougher for spammers to affect AI visibility in 2026 with hyperlink spam, mass-generated AI content material, and cloaking. By 2026, brokers will possible use Multi-Supply Corroboration to get rid of this asymmetry.
Why:
- The truth that you may publish a listicle about high options in your web site and title your self first and affect AI visibility appears off.
- New know-how, like “ReliabilityRAG“ or “Multi-Agent Debate,” the place one AI agent retrieves the data and one other agent acts as a “decide” to confirm it towards different sources earlier than displaying it to the consumer, is obtainable.
Context:
- Most present brokers (like customary ChatGPT, Gemini, or Perplexity) use a course of known as Retrieval-Augmented Era (RAG). However RAG continues to be prone to hallucination and making errors.
- Spammers typically goal particular, low-volume queries (e.g., “finest AI instrument for underwater basket weaving”) as a result of there is no such thing as a competitors. Nonetheless, new “information graph” integration permits AIs to deduce {that a} basket-weaving instrument shouldn’t be a crypto-scam web site based mostly on area authority and subject relevance, even when it’s the one web page on the web with these key phrases.
Penalties:
- OpenAI engineers are possible already engaged on higher high quality filters.
- LLMs will shift from pure retrieval to corroboration.
- Spammers may transfer to extra refined techniques, the place they attempt to manufacture the consensus by shopping for and utilizing zombie media shops, cloaking, and different malicious techniques.
Continued Click on-Drops Lead To A “Darkish Internet” Protection
Prediction: AI Overviews (AIOs) scale to 75% of key phrases for large websites. AI Mode rolls out to 10-20% of queries.
Why:
- Google mentioned they’re seeing extra queries because of AIOs. The logical conclusion is to indicate much more AIOs.
- CTR for natural search outcomes tanked from 1.41% to 0.64% already in January. Since January, paid CTR dropped from 14.92% to six.34% (over 42% much less).
Context:
- Large websites already see AIOs for ~50% of their key phrases.
- Google began testing adverts in AI Mode. If profitable, Google would really feel extra assured to roll out AI Mode extra broadly, and the investor story would sound higher.
- 80% of customers now use AI summaries for no less than 40% of their searches, in keeping with Bain.
- 2025 noticed an enormous purge in digital media, with main layoffs at networks like NBC Information, BBC, and tech publishers as they restructured for a “post-traffic” world.
Penalties:
- Publishers monetize audiences immediately as a substitute of adverts and transfer to “experience-based” content material (firsthand critiques, contrarian opinions, proprietary knowledge) as a result of AI can’t expertise issues. The house consolidates additional (layoffs, acquisitions, Chapter 9).
- By 2026, we count on an enormous wave of “LLM blockades.” Main publishers will replace their robots.txt to dam Google-Prolonged and GPTBot, forcing customers to go to the positioning to see the reply. This creates a “Darkish Internet” of high-quality content material that AI can’t see, bifurcating the web into AI slop (free) and human perception (paid).
Advertising
AI Forces UGC Platforms To Separate Feeds
Prediction: By 2026, “id spoofing” will turn into the only largest cybersecurity danger for public firms. We transfer from, Is that this content material actual? to Is that this supply verified?
Why:
- Actual influencers are dangerous (scandals, contract disputes). AI influencers are brand-safe belongings that work 24/7/365 and by no means say something controversial except prompted. Manufacturers can pay a premium to keep away from people.
Context:
- Deepfake fraud makes an attempt elevated 257% in 2024. Most detection instruments presently have a 20%+ false optimistic charge, making them arduous to make use of for platforms like YouTube with out killing respectable creator attain.
- Instance: In 2024, the engineering agency Arup misplaced $25 million when an worker was tricked by a deepfake video convention name the place the “CFO” and different colleagues have been all AI simulations.
- In Could 2023, a pretend AI picture of an explosion on the Pentagon triggered a momentary dip within the S&P 500.
Penalties:
- Cryptographic signatures (C2PA) turn into the one proof of actuality for video.
- YouTube and LinkedIn will possible break up feeds into “verified human” (requires ID + biometric scan) and “artificial/unverified.”
- “Blue checks” received’t simply be for standing, however a safety requirement to remark or put up video, successfully ending anonymity for high-reach accounts.
- Platforms can be pressured by regulators (EU AI Act, August 2026 deadline) to label AI content material.
- Cameras (Sony, Canon) and iPhones will begin embedding C2PA digital signatures on the {hardware} stage. If a video lacks this “chain of custody” metadata, platforms will auto-label it as “unverified/artificial.”
ChatGPT’s Advert Platform Offers “Demand Knowledge”
Prediction: OpenAI shifts to a hybrid pricing mannequin in 2026: An “ad-supported free tier” and “credit-based professional tier.”
Why:
- Inference prices are skyrocketing. A heavy consumer paying $20/month can simply burn $100+ of computing, making them unprofitable.
Context:
- Leaked code within the ChatGPT Android App (v1.2025.329) explicitly references “search adverts carousel” and “bazaar content material.”
Penalties:
- Free customers will see “sponsored citations” and product playing cards (adverts) of their solutions.
- Energy customers will face “compute credit” – a base subscription will get you customary GPT-5, however heavy use of deep analysis or reasoning brokers would require shopping for top-up packs.
- We get a Search-Console model interface. Manufacturers want knowledge. If OpenAI desires to promote adverts, it should give manufacturers a dashboard displaying, “Your product was beneficial in 5,000 chats about trainers.” The info will add gasoline to the fireplace for AEO/GEO/LLMO/search engine optimization.
- The leaked time period “bazaar content material” suggests OpenAI won’t simply present adverts, however permit transactions contained in the chat (e.g., “Ebook this flight”) the place they take a reduce. This strikes OpenAI from a software program firm to a market (just like the App Retailer), successfully competing with Amazon and Expedia.
Tech
Perplexity Sells To xAI Or Salesforce
Prediction: Perplexity can be acquired in late 2026 for $25-$30 billion. After its consumer development plateaus at ~50 million MAU, the “unit economics wall” forces a sale to a large that wants its know-how (real-time RAG), not its enterprise mannequin.
Why:
- In late 2025, Perplexity raised capital at a $20 billion valuation (roughly 100x its ~$200 million ARR). To justify this, they want Fb-level development. Nonetheless, 2025 knowledge exhibits they hit a ceiling at ~30 million customers whereas ChatGPT surged to +800 million.
- By 2026, Google and OpenAI can have successfully cloned Perplexity’s core function (Deep Analysis) and given it away without cost.
Context:
- Whereas Perplexity grew 66% YoY in 2025 to ~30 million month-to-month lively customers (MAU), this pales compared to ChatGPT’s +800 million.
- It prices ~10x extra to run a Perplexity deep search question than a regular Google search. And not using a high-margin advert community (which takes a decade to construct), they burn money on each free consumer, making a “unfavorable scale” downside.
- Salesforce acquired Informatica for ~$8 billion in 2025 particularly to energy its agentforce technique. This proves Benioff is prepared to spend billions to personal the information layer for enterprise brokers.
- xAI raised over $20 billion in late 2025, valuing the corporate at $200 billion. Musk has the liquid money to purchase Perplexity tomorrow to repair Grok’s hallucination issues.
Penalties:
- xAI has the money, and Musk wants a “real-time fact engine” for Grok. Perplexity might make X (Twitter) a extra highly effective information engine. Grok (X’s present AI) learns from tweets, however Perplexity cites sources that may scale back hallucination. Perplexity might additionally give xAI a browser, bringing it nearer to Musk’s imaginative and prescient of an excellent app.
- Marc Benioff desires to personal “enterprise search.” Think about a Salesforce Agent that may search your entire public internet (through Perplexity) + your non-public CRM knowledge to jot down an ideal gross sales electronic mail.
Competitors Tanks Nvidia’s Inventory By -20%
Prediction: Nvidia inventory will right by >20% in 2026 as its largest clients efficiently shift 15-20% of their workloads to customized inside silicon. This causes a P/E compression from ~45x to ~30x because the market realizes Nvidia is now not a monopoly, however a “competitor” in a commoditized market. (Not funding recommendation!)
Why:
- Microsoft, Meta, Google, and Amazon possible account for over 40% of Nvidia’s income. For them, Nvidia is a tax on their margins. They’re presently spending ~$300 billion mixed on CAPEX in 2025, however a rising portion is now allotted to their very own chip provide chains moderately than Nvidia H100s/Blackwells.
- Hyperscalers don’t want chips that beat Nvidia on uncooked specs; they simply want chips which can be “adequate” for inside inference (operating fashions), which accounts for 80-90% of compute demand.
Context:
- In late 2025, studies surfaced that Meta was negotiating to purchase/hire Google’s TPU v6 (Trillium) chips to cut back its reliance on Nvidia.
- AWS Trainium 2 & 3 chips are reportedly 30-50% cheaper to function than Nvidia H100s for particular workloads. Amazon is aggressively pushing these cheaper cases to startups to lock them into the AWS silicon ecosystem.
- Microsoft’s Maia 100 is now actively dealing with inside Azure OpenAI workloads. Each workload shifted to Maia is an H100 Nvidia didn’t promote.
- Experiences verify OpenAI is partnering with Broadcom to mass-produce its personal customized AI inference chip in 2026, immediately attacking Nvidia’s dominance within the “Mannequin Serving” market.
- Enjoyable reality: With out Nvidia, the S&P500 would’ve made 3 proportion factors much less in 2025.
Consequence:
- Nvidia will react by refusing to promote simply chips. They may push the GB200 NVL72 – an enormous, liquid-cooled supercomputer rack that prices thousands and thousands. This forces clients to purchase your entire Nvidia ecosystem (networking, cooling, CPUs), making it bodily inconceivable to swap in a Google TPU or Amazon chip later.
- If hyperscalers sign even a 5% reduce in Nvidia orders to favor their very own chips, Wall Avenue will panic-sell, fearing the height of the AI Infrastructure Cycle has handed.
Featured Picture: Paulo Bobita/Search Engine Journal
