Friday, June 20, 2025

A World Where "Answers" are the Issue, Not "Search" Results

The replacement of traditional search with language model “answers” shifts the internet from a link-based content ecosystem to a world where traffic is less critical than relevance and authority. 


Reduced organic traffic is already happening. 


Study/Source

Content Provider

Timeframe

Reported Traffic Decline

Notes/Findings

LinkedIn (Poffel, 2025) 1

HubSpot (Marketing blog)

Mar 2023–Jan 2025, Google AI/SGE

75% decrease (24.4M → 6.1M)

Drop attributed to AI-generated answers reducing user clicks; debate over Google updates vs AI.


Chegg (Education Q&A)

2024–2025, Google AI Overviews

34% decrease (5.6M → 3.7M)

Chegg sued Google, alleging AI Overviews use their content to keep users on Google.


Stack Overflow (Programming Q&A)

2024–2025, AI tools (ChatGPT, Copilot)

Significant decline (unspecified)

Developers get instant answers from AI, reducing visits to Stack Overflow.


Informational sites (various)

SGE early tests

18–64% decrease

Especially for "easily answerable" informational queries.

The Hoth (via MarketingEdge) 2

General publishers (featured in AI Overview)

Post-May 2024, Google AI Overviews

8.9% average drop

Sites not featured: 2.6% drop. Some small publishers: up to 70% decline (Bloomberg report).

Forbes (2025) 3

Industry-wide

Google AI Overviews

15–64% decline

60% of searches now yield zero clicks; top links pushed down, reducing click-through rates.

Bain & Company (2025) 4

General consumer search

2025, AI summaries in search

15–25% decrease

80% of users rely on AI summaries for 40%+ of searches; 60% of searches end without a click.

BrightEdge, SurferSEO, Conductor (via WhistlerBillboards) 5

Mail Online (news), general sites, bloggers

March–May 2025, Google AI Overviews

Mail Online: 56% CTR drop; SurferSEO: 34.5% CTR drop for position 1; Conductor: up to 60% traffic drop

Fashion, travel, DIY, cooking, tech review, and health bloggers report up to 70% traffic loss.

Ahrefs (via WhistlerBillboards) 5

Small recipe and health bloggers

2024–2025, Google AI Overviews

Up to 65% of top-page traffic lost

“How to” and “what is” queries especially impacted.

Chegg (via WhistlerBillboards) 5

Chegg (Education Q&A)

Jan 2024–Jan 2025

49% decline in non-subscriber traffic

Attributed to AI Overviews.


Websites dependent on high-volume, low-depth traffic arguably are at risk, as the chatbots can aggregate this information without linking back.


On the other hand, language models prioritize content that is authoritative, unique, or contextually rich. That doesn’t always or necessarily mean a citation, but might be the necessary precondition. 


The models likely will favor content from established experts or primary sources as well.


So content creators may need to optimize for being "noticed" and referenced by language models,  rather than ranking high in search results.


Thursday, June 19, 2025

Stable coins for Walmart, Amazon?

Walmart and Amazon are said to be exploring the creation of their own stablecoins,  cryptocurrencies whose value is pegged to that of another currency, commodity, or financial instrument such as gold, Treasury bills or the U.S. dollar.


For those of us who do not buy and sell crypto as an investment or store of value, such stablecoin payment methods are a practical application of crypto in our daily lives.


New proposed legislation passed by the U.S. Senate to create a regulatory framework for stablecoins arguably will help clear the way, pegging stablecoin value to the U.S. dollar. The U.S. House of Representatives already has passed similar legislation. 


Most of the value seems tied to the ability to reduce interchange payments (usage fees, essentially) paid to credit card processors Visa and Mastercard, for example. By leveraging stablecoins, they can process payments at lower cost, with faster settlement, and maintain greater control over their transaction cost infrastructure. 


Feature

Credit Card Payments

Stablecoin Payments

Typical Fee

2-3%+ per transaction

1.5% or lower

Settlement Time

1-3 business days

Real-time or near real-time

Network Intermediaries

Banks, Card Networks

Minimal (blockchain only)

Retailer Control

Low

High

Volatility Risk

None (fiat)

Low (fiat-pegged)


Such retailer stablecoins might be considered a major boost for use of cryptocurrencies as a payment mechanism or currency. Over time, if consumers embrace the method, it also has implications for the fortunes of credit card processing networks as well as the major retailers who use the stablecoins. 


Banks presumably also will have to adjust, as they are the processing network partners and actually issue credit cards, authorize and receive payments consumers make with retailers.


Wednesday, June 18, 2025

Will AI Disrupt "Routine" or "Complex" Business Functions and Job Roles?

One of the emerging paradoxes of artificial intelligence is that its greatest value for business processes includes both processes that are “routine” or “repetitive,” or, on the other hand, complex. 


But the key lies in relative value. High-value AI use cases typically involve complex decision-making, pattern recognition, predictive analytics, or end-to-end process automation. 


Lower-value use cases are often repetitive, routine, or administrative tasks where AI provides incremental, but not transformative, improvements. 


So even if some might think AI will be most disruptive for lower-skill occupations and tasks, it seems highly likely that the bigger amount of disruption will come for job functions that are cognitive, across most industries. 


Industry

High-Value AI Use Cases

Lower-Value AI Use Cases

Healthcare

Diagnostics (image/lab analysis), personalized treatment, drug discovery, operational optimization 1,2,3

Routine admin tasks (e.g., appointment reminders), basic data entry

Finance

Fraud detection, algorithmic trading, credit risk assessment, automated loan processing 1,2

Simple customer queries, basic transaction categorization

Manufacturing

Predictive maintenance, quality control, supply chain optimization 2,3

Standard inventory updates, basic production scheduling

Retail

Personalized recommendations, demand forecasting, dynamic pricing, customer support bots 2,3

Standard order processing, generic marketing emails

Software Development

Code generation, intelligent code completion, automated testing/debugging, documentation generation 4,5

Simple script writing, basic code formatting

Education

Adaptive learning platforms, personalized content, automated grading 1,3

Scheduling classes, distributing standard materials

Business Operations

Hyperautomation (end-to-end process automation), advanced analytics, generative content for marketing/HR 6 3

Expense report filing, simple workflow notifications


Tuesday, June 17, 2025

AI Inference Costs Will Become More Predictable, as Did Cloud Computing "As a Service" Costs

Though it is rational to note that AI inference costs are somewhat unpredictable at the moment, that also was true of cloud computing in general. But as the technology matures, it is likely--probably inevitable--that AI inference “as a service” will develop methods of providing customers more cost predictability. 

 

Cloud computing, after all, also featured unpredictable, often expensive usage-based pricing, that made customer budgeting difficult. 


But the market responded, creating pricing models (reserved instances, committed use discounts, and hybrid approaches) that provided cost predictability.


AI inference is likely following a similar trajectory. Token-based pricing means the cost per inference can fluctuate based on model complexity, input length, and provider capacity.


But providers already are experimenting with different approaches beyond pure pay-per-token models: subscription tiers, reserved capacity options or volume discounts that provide more predictable monthly costs. 


Enterprise contracts increasingly include committed usage terms that offer better rate predictability as well. 


And  competition also will drive providers to offer more customer-friendly pricing structures. AWS, Google Cloud, and Azure all evolved toward more predictable pricing options as the market matured and customers demanded better cost management tools.


At the same time,  models, hardware and inference acceleration will naturally drive down costs. So will analytics. 


Some of us cannot see the cost unpredictability being a long-term issue for AI inference.


A World Where "Answers" are the Issue, Not "Search" Results

The replacement of traditional search with language model “answers” shifts the internet from a link-based content ecosystem to a world where...