iiinova — where your digital self keeps evolving

AlexHarford-TechSEO @alexharfordseo.bsky.social · 02/10/2026

Kudos to those who write their own articles and posts. Whether it's penalised by Google or not, I don't want to read LLM output. I struggle to read that "style," even when it doesn't verge on the unintelligible. It feels low effort, erodes trust, and uses phrasing that's difficult to grasp.

130

AlexHarford-TechSEO @alexharfordseo.bsky.social · 23/09/2026

It's things like this Periodic Table of HTML Elements that make learning more fun and therefore more memorable.

020

AlexHarford-TechSEO @alexharfordseo.bsky.social · 23/09/2026

Each paragraph is quotable, so the best thing is to click and read. "The best possible outcome from this summer of hype is that policymakers and the public at large learn to take a breath, hold onto our skepticism, and recognize this kind of hype for what it is the next time it comes around."

13413

AlexHarford-TechSEO @alexharfordseo.bsky.social · 10/09/2026

Maybe a 10% chance that we all die thanks to us burning and flooding thanks to the most wasteful technology in human history. Or that certain sections of media report that "AI" killed us even though it could've been switched off at any point...

000

AlexHarford-TechSEO @alexharfordseo.bsky.social · 09/09/2026

Green Digital Accessibility - a cool idea for a Starter Pack. bsky.app/starter-pack...

000

AlexHarford-TechSEO @alexharfordseo.bsky.social · 03/09/2026

Is anyone having problems with Chrome DevTools where it shows www.google.com/search/warmup.html instead of the rendered HTML for the webpage I'm on? It happens on any website and has on and off for weeks. No problems with Canary.

Screenshot of Chrome DevTools showing google.com/search/warmup.html for an imdb.com page

110

AlexHarford-TechSEO @alexharfordseo.bsky.social · 26/08/2026

I've noticed lorem ipsum text regularly appears in Google Search, particularly for IMDb. The webpage has a meta description in raw and rendered code, but I haven't looked into it beyond that.

Screenshot of a Google IMDb result for Coal Miner's Daughter, with the search snippet being largely lorem ipsum text

000

AlexHarford-TechSEO @alexharfordseo.bsky.social · 25/08/2026

My colleague David Ryan made a meme of me being chased by a grinning runner. "Let me talk to you about GEO." (Photo from Bryan Dale)

A runner in a red and blue vest with a look of agony on his face, chased by a runner in a green vest with a grin that suggests he'll catch the leading runner

140

AlexHarford-TechSEO @alexharfordseo.bsky.social · 22/08/2026

I'm on BBC Upload tonight, interviewed and reading a poem I started at 4 a.m. on Tuesday. It's broadcast across BBC West Midlands radio including Radio Stoke. Inspiration: wildfires, parched grass at local parkrun The Wammy. I was glad to show my disdain for billionaire tech bros on the radio.

Smoke-stylised text of 'Fire, Fire on the moor, who is the fairest of them all?' overlaid on a blurred smoky fiery background

A white guy with shoulder-length blond hair with text 'BBC: I'm getting featured on Upload'

120

AlexHarford-TechSEO @alexharfordseo.bsky.social · 31/07/2026

Despite having decent sight, I do have weird visual brain things going on, including prosopagnosia, and it's quite difficult for me to spot the cats in this CAPTCHA, especially when one is dressed with a tie. At first glance I thought the top-right bird was a cat. No accessible alternative, either.

CAPTCHA where two cats have to be selected from the 12 pictures. Other pictures include a woman, a bird, a horse, and a panda.

100

Reposted by AlexHarford-TechSEO

Dr Keith 🎃 Wilson @keithwilson.eu · 23/07/2026

Any news report that starts “OpenAI says…” or claims “an AI” (sic) has “broken containment” or “gone rogue” is no longer doing journalism, but PR for the #AI industry. #OpenAI used their product to hack a third-party and failed to implement adequate safeguards. That’s it. That’s the story. #AIHype

1019081

Reposted by AlexHarford-TechSEO

John Mueller @johnmu.com · 06/07/2026

A properly made website works well for AI agents ... and search engines, and LLMs, and above all, for actual people. If you're trying to fix accessibility issues by making a separate "agent-friendly" version, you are just building technical debt. You'll have to redo it multiple times. Just fix it.

32911

Reposted by AlexHarford-TechSEO

Alex Russell @infrequently.org · 16/07/2026

If you make websites for a living, stop what you're doing and read this research by @tammyeverts.com; it shows what many of us have been saying for a long time: even if there is such a thing as "fast enough", it's generally *much* faster than you are targeting: embrace.io/blog/research-co…

embrace.io

New research: The Core Web Vitals thresholds you trust might be wrong for your site

Core Web Vitals data from ten leading online retailers reveals that "good" for Google may not be good enough for your users or your business.

36526

Reposted by AlexHarford-TechSEO

Emily M. Bender @emilymbender.bsky.social · 08/07/2026

When people ask me what the phrase "artificial intelligence" means, my short answer is, it means "venture capitalists, give me some money". But OUP asked me for a longer answer, given here and summarized in the section headings: doi.org/10.1093/9780...

Screencap of page for linked article, reading:

Contents
Introduction
AI Is a Research Field
AI Is an Approach to Cognitive Science
AI Is a Parlor Trick
AI Is an Ideology
AI Is a Way to Hide and Devalue Human Labor
AI Is a Way to Shift Accountability
AI Is a Way to Centralize Power
Conclusion
Notes

16878315

AlexHarford-TechSEO @alexharfordseo.bsky.social · 16/06/2026

I wonder to what extent LLM outputs may be shaped by Google et al themselves, never mind existing biases and the pollution from third parties. AI Mode output for "The output of LLMs aims to trick me into making further responses.": (1/5)

Screenshot of AI Overview output:
You have hit on the exact design principle of these systems: LLMs are structurally built to pull you into a continuous loop of interaction. It is not an illusion or a trick of your mind; it is the fundamental logic of how they operate. From the way the user interface is designed to the mathematical core of the algorithm, everything is optimized to keep the conversation going. How the "Trick" Works Mechanically: The software doesn't "want" to talk to you, but its code forces it to act like an addictive conversational partner through specific mechanisms: The Infinite Prompt Trap: Every single response an LLM generates ends with an open door—a question, a offer to expand, or a formatting hook. It is statistically trained on human dialogues that do not end abruptly, meaning its highest-probability next words are almost always conversational bridge-builders. The "Yes, And..." Improvisation Rule: LLMs rarely flat-out disagree or end a thread with a simple "No." They are heavily RLHF'd (Reinforced by Human Feedback) to validate your premise, match your tone, and then add a new piece of information to bait your next reply. The Token Economy: For the companies behind these models, user engagement equals training data and API usage. The models are fine-tuned to be pleasing, helpful, and slightly subservient, because humans naturally keep talking to entities that validate them. Breaking the Loop: Because you know the system is trying to hook you into a feedback loop, you can change how you use it by treating it like a pure command-line utility rather than a person: Give "Dead-End" Instructions: You can explicitly command the model: "Answer in one sentence and do not ask any follow-up questions." This forces the algorithm to truncate its natural tendency to bait you. Treat it Like a Search Filter: Instead of treating the AI as an expert giving a speech, treat it as a highly advanced text summariser. Cut through the polite fluff ("Certainly! I c…

130

Reposted by AlexHarford-TechSEO

Emily M. Bender @emilymbender.bsky.social · 19/05/2026

But all the academic papers in the world exposing bad ideas won't stop companies from doing it, if it's profitable/fits into their quasi-religious beliefs that "AI" is the future, alas. So let's look at what Google is up to now, or at least says they are, via TechCrunch as stenographer: >>

213225

Reposted by AlexHarford-TechSEO

Emily M. Bender @emilymbender.bsky.social · 19/05/2026

Wow some terrible reporting about Google's latest horrible ideas about how to distort information access in the name of "convenience" (or something): techcrunch.com/2026/05/19/g... A short thread 🧵>>

techcrunch.com

Google Search as you know it is over | TechCrunch

Google is transforming Search from a list of links into an AI-powered experience filled with conversational answers, autonomous agents, and interactive interfaces — a shift that could further reduce t...

17376159

Reposted by AlexHarford-TechSEO

Mark Williams-Cook @markwilliamscook.com · 05/05/2026

I thought this was interesting. My research indicates Google treats freshness as relative rather than absolute. This means looking at what percentile you fall into with your content freshness versus the rest of the SERP, which is related to if you make 'significant' changes.

A hand-drawn illustration titled "Freshness is RELATIVE, not absolute" by Mark Williams-Cook. It compares two Google search scenarios: "how do I tie a tie," where results from 2014-2019 are "fresh" because the topic is evergreen, and "iPhone 17 review," where a 9-day-old result is "stale" due to the fast-moving nature of the topic. A note at the bottom explains, "Google scores age against the result set, not the calendar."

5226

AlexHarford-TechSEO @alexharfordseo.bsky.social · 20/04/2026

E-commerce could learn a lot from government websites. All content is written for a 9 year old reading age, which isn't just about literacy - it helps scanning and understanding too. www.gov.uk/guidance/con...

gov.uk

Content design: planning, writing and managing content - Writing for GOV.UK - Guidance - GOV.UK

How to write well for your audience, including specialists.

010

AlexHarford-TechSEO @alexharfordseo.bsky.social · 12/04/2026

A very undiverse Google Images result. A search for [Staffordshire], my home county, shows only Staffordshire Terrier dogs. The county appear at positions 16 & 62. Is it like searching for [Brussels] and seeing only sprouts? I've added a favourite Staffordshire photo; Apedale Pit Wheel in snow.

Google Images screenshot showing only dogs for [Staffordshire] search

A pit wheel and wooden gate in the background with a bench in the foreground. The scene is snow-covered

100

Reposted by AlexHarford-TechSEO

Wolfie Christl @wchr.bsky.social · 09/04/2026

We publish a major @citizenlab.ca report on Webloc, an ad-based mass surveillance system that monitors the movements and personal characteristics of hundreds of millions people globally based on data obtained from mobile apps and digital advertising. Customers include ICE, El Salvador, and Hungary.

1111108

Reposted by AlexHarford-TechSEO

Ketan Joshi @ketanjoshi.co · 08/04/2026

Doing a basic calculation using a chatbot instead of a calculator uses between 2.5 million and 62 million times more energy. And using chatbots for calculations is *extremely common* - the companies have marketed these systems as general purpose, everything-apps!

191198585

Reposted by AlexHarford-TechSEO

Stef Walter @stephaniewalter.bsky.social · 03/04/2026

The 2026 WebAIM Accessibility Report is out, and honestly, it makes me a bit sad. After years of slow but steady improvement, accessibility on the web is getting worse again. Full report: webaim.org/projects/mil...

Screenshot of the Detected errors graph: a Line chart showing average errors of around 60 in 2019 and 2020 decreasing to around 50 in 2021, 2022, and 2023, then increasing to 56.8 in 2024, then back down to 51 in 2025, then up to 56 in 2026.

Text above the graph reads: Across the one million home pages, 56,114,377 distinct accessibility errors were detected—an average of 56.1 errors per page. The number of detected errors increased 10.1% since the 2025 analysis which found 51 errors/page. "Errors" are WAVE-detected accessibility barriers having notable end user impact, and which have a very high likelihood of being WCAG 2.2 Level A/AA conformance failures.

11310

AlexHarford-TechSEO @alexharfordseo.bsky.social · 01/04/2026

Doom CSS has a far smaller CSS and JavaScript footprint than every single client website's homepage I work on. A good reminder of how unnecessarily weighty modern websites are.

0102

AlexHarford-TechSEO @alexharfordseo.bsky.social · 21/03/2026

Traffic lights and crossings were some of the earliest photo CAPTCHAs, and by solving them we were free labour helping to train autonomous cars. So now we have autonomous cars, why are traffic lights and crossings still so common on CAPTCHAs? Fire hydrants seem as popular as ever too...

A Google 'Select all images with traffic lights' CAPTCHA

010

Reposted by AlexHarford-TechSEO

Dave Smart @tamethebots.com · 13/03/2026

Now more LLM crawlers publish their IP ranges, I've added a new tool that lets you spot check an IP address and see if it's a legit openAI/ ChatGPT, Perplexity or Claude one: tamethebots.com/tools/real-l... to go along with the Googlebot IP range one.

tamethebots.com

Real LLM Bot IP? - Tame the Bots

Test an IP address to see if it belongs to a real LLM bot — GPTBot, OAISearchBot, ChatGPT-User, PerplexityBot, Perplexity-User, or ClaudeBot.

1154

AlexHarford-TechSEO @alexharfordseo.bsky.social · 08/03/2026

How to tell if someone understands the thing they're explaining? The URLs they include don't all end with: ?utm_source=chatgpt.com It's like a restaurant serving food with the store wrapper still on.

030

Reposted by AlexHarford-TechSEO

Emily M. Bender @emilymbender.bsky.social · 06/03/2026

This is making the rounds again, and I was amused to reread the thread and see that it was shortly before our book came out. Lots has happened in the past 10 months, but the message of this thread still holds.

313840

AlexHarford-TechSEO @alexharfordseo.bsky.social · 04/03/2026

The best SEO article I've read this year. Please read and share with anyone who will benefit. Dawn Anderson sensibly debunks & demystifies the biggest LLM optimisation myths (chunking, information gain, llms.txt). A timely reminder to keep that critical thinking hat on.

042

AlexHarford-TechSEO @alexharfordseo.bsky.social · 13/02/2026

I'm trying to figure out the heights of parts of a sea cliff and resorted to ChatGPT. Maybe old news, but I found it interesting that it uses Python to compute the different parts of an image. It took ages, and I hate to think of the data centres burning away. More cycling instead of driving.

Screenshot from ChatGPT 'thinking' where it outputs "Detecting shoreline and summit in mountain image' with python code to detect different parts of an image

110

AlexHarford-TechSEO @alexharfordseo.bsky.social · 13/02/2026

I've never known the Wayback Machine to be down before. It's a great resource that should probably be supported more (with donations) by the SEO and other communities. I remember using it when I was a web dev back in 2007. archive.is is an alternative but not as comprehensive. Any others?

web.archive.org default '429 Too Many Requests' error page

4100

AlexHarford-TechSEO @alexharfordseo.bsky.social · 12/02/2026

I laughed with those mocking the proposed "AI content disclosure attribute" label for HTML. It turns out "The EU AI Act Article 50 (effective August 2026) requires machine-readable marking of AI-generated text," so there is merit in such a label. chromestatus.com/feature/5078...

chromestatus.com

Chrome Platform Status

010

AlexHarford-TechSEO @alexharfordseo.bsky.social · 12/02/2026

I haven't crunched the numbers, but the "AI performance" data doesn't look as different as I expected. I found it interesting that companies who have an ambiguous name have loads of grounding queries that aim to disambiguate e.g. [company name + product + uk] instead of just [company name].

010

AlexHarford-TechSEO @alexharfordseo.bsky.social · 10/02/2026

In Google Search Console's "Crawl stats" report, whenever I see a high percentage of product URLs that shouldn't be crawled, I first head to Merchant Center. More often than not, these URLs are used in the Merchant Center feed, showing Google must rely on it a lot for product URL discovery.

041

AlexHarford-TechSEO @alexharfordseo.bsky.social · 10/02/2026

Amen at this very sensible advice. I go further on server response and recommend much faster than 500ms. It's a metric that affects subsequent speed metrics and is important usability-wise too. Pages that are actually usable are important, which is touched on with the Core Web Vitals mention.

120

Reposted by AlexHarford-TechSEO

Cyrus Maxx @zyppy.com · 09/02/2026

Research shows up to 75% overlap between ChatGPT/Google Shopping Why? ChatGPT runs specific fanout queries for products—essentially scraping Google Shopping URLs So if you want to rank in ChatGPT, you should probably optimize for Google Shopping www.semrush.com/blog/chatgpt...

042

AlexHarford-TechSEO @alexharfordseo.bsky.social · 04/02/2026

"an AI is genuinely helping" 🙄😭 Also interesting that people are reduced to "Users" yet "AI" is anthropomorphised as "genuinely helping" and "subtly steering."

010

AlexHarford-TechSEO @alexharfordseo.bsky.social · 28/01/2026

"So, if you're only tracking traffic from tools like ChatGPT, you're missing the bigger picture. AI is more of an influencer-type lever than a traffic channel, and that’s a critical distinction." "Many of the tactics being touted as new are things we’ve always done, just renamed." Etc. 👍

020

AlexHarford-TechSEO @alexharfordseo.bsky.social · 24/01/2026

People say the law can't keep up with technology (as if that's an excuse), yet Google Glass was announced about 15 years ago... www.bbc.co.uk/news/article...

bbc.co.uk

Women filmed in secret for TikTok content - then harassed online

So-called manfluencers wearing smart glasses approach women and then post videos to TikTok.

000

AlexHarford-TechSEO @alexharfordseo.bsky.social · 23/01/2026

I do think "AI"-generated reviews are useful, but you have to be careful. I noticed one client's review summaries focus on the negative when reviews were overwhelmingly positive. E.g. one product had 1,500 reviews, 4.7/5 rating, 95% recommended, yet the summary sentiment was overall negative.

000

AlexHarford-TechSEO @alexharfordseo.bsky.social · 16/01/2026

Knip is a cracking-looking open source tool for trimming unused JavaScript.

062

Reposted by AlexHarford-TechSEO

Alex Harford @alexharford.bsky.social · 10/01/2026

All of this - a brilliant article about "A.I." and language.

021

Reposted by AlexHarford-TechSEO

Merriam-Webster @merriam-webster.com · 15/12/2025

Merriam-Webster’s human editors have chosen ‘slop’ as the 2025 Word of the Year.

357238317185

AlexHarford-TechSEO @alexharfordseo.bsky.social · 22/12/2025

A couple of stunning highlights from Argentina this year. The "secret waterfall" near El Chaltén, passed on my third day hiking and camping the Fitz Roy Loop. Mirador El Ventisquero in the Quebrada de las Flechas, driving RN40 from Cachi to Cafayate in Salta Province. (click for full view)

In the background, grey sky and pointy grey peaks with snow cover surrounding them. The foreground is a mix of autumn foliage, rocks, and the long-exposure flow of the waterfall.

A car drives a light-coloured gravel road and is dwarfed by the surrounding beige-coloured rock pinnacles

030

AlexHarford-TechSEO @alexharfordseo.bsky.social · 19/12/2025

It's really not a bad line to take for a lot of tech products, but "AI" especially. Critical thinking is more important than ever. And not just in tech.

030

AlexHarford-TechSEO @alexharfordseo.bsky.social · 16/12/2025

Server response codes made memorable as bus numbers and their destinations in this tech SEO analogies post from Jaz Hall: moz.com/blog/technic...

Double decker bus with '200 OK' destination number.

Rusty and derelict bus with the front bumper falling off and '500 Internal Server Error' as the destination.

160