---
title: "The New Host My Blog, And Why Search Has Split In Two"
url: https://hostmy.blog/we-rebuilt-host-my-blog/
date: 2026-08-16
modified: 2026-09-03
lang: en
author: "Aditya Sharma"
description: "The new site is live. Here is what changed, and why ranking on Google is no longer the same job as being cited by an AI answer engine."
categories:
  - "AI Search"
  - "Announcement"
image: https://hostmy.blog/wp-content/uploads/2026/08/og-1437-the-new-host-my-blog-1024x538.png
word_count: 2113
---

# The New Host My Blog, And Why Search Has Split In Two

The new Host My Blog is live.

I have wanted to build this for a long time. The old site was not broken. It just described a company we stopped being years ago, and every month I left it alone was another month we explained the real work over email.

So we rebuilt it. New look, new structure, more honesty about the job.

The redesign is the small half of this post. The bigger half is what changed underneath it, and where we are taking this next.

## What changed

Search split in two while most of us were still optimising for one.

For most of the years I have done this, it was one job. Write something worth reading. Keep the server fast. Clear the crawl errors. Google made a list, a person read it and clicked. That machine still runs and still pays the bills.

A second machine now runs beside it. Somebody asks ChatGPT or Perplexity, or Google puts an AI Overview above everything, and no list gets made. A model reads a few documents, writes one answer, credits two or three sources underneath.

**Being fourth in a list of ten is a real position with real traffic. Being fourth candidate for an answer that cites three is nothing.**

One question, two machines

ONE QUESTION, TWO MACHINES

THE LIST
Google Search
Ten blue links.
A person picks one.

-
IT READS
Your HTML
Your speed
Your links

YOU FINISH FOURTH
You are on the page.
You still get clicks.

THE ANSWER
AI Answers
AI Overviews, ChatGPT, Perplexity.
One answer, about three citations.

IT READS
Your structure
Your clarity
Your first-hand experience

YOU FINISH FOURTH
You are not in the answer.
You get nothing.

Both machines are running right now.
The same blog has to be legible to both, and only one of them tells you how it went.

## The numbers, honestly

I do not enjoy throwing statistics around, so here is one you can go and read yourself.

Pew Research Center tracked the browsing of 900 US adults through March 2025 and [published it in July 2025](https://www.pewresearch.org/short-reads/2025/07/22/google-users-are-less-likely-to-click-on-links-when-an-ai-summary-appears-in-the-results/). Three numbers matter to you.

**8% versus 15%.** Where an AI summary appeared, people clicked a search result in 8% of visits. Where none appeared, 15%. Close to half.
- **1%.** That is how often they clicked a link inside the summary itself.
- **26% versus 16%.** Visits that ended the browsing session entirely, with a summary and without.

Be careful with those. One month, one country, Google only, 900 people. That is not your traffic. But it was measured rather than asserted, which is more than most of what you will read on this.

Two more from the same study, and these are the useful ones. **18% of all searches produced an AI summary, but 60% of searches starting with a question word did.** And 8% of two word searches, against 53% of searches ten words or longer.

Read that as: the more your reader types like a human asking a question, the more likely they never reach you.

## Who this hits first

If you sell something, a lost click is a lost chance at a sale. Annoying, survivable.

If you run ads, a lost click is the product.

**Ad-funded publishers get billed for this twice over, because sessions are not how they measure the business. Sessions are the business.**

That is the part I keep repeating to people, because it is easy to miss. A Mediavine-scale publisher has already done the hard work. Real recipes, real reviews, real years of it. The revenue arrives as a session, an impression against that session, and a rate on that impression. Break the first link in that chain and the other two never happen. There is no second funnel underneath to catch it.

So when clicks move from a list into an answer, a publisher on an ad model does not lose a slice of the upside. They lose the whole unit.

Bloggers just starting out and publishers running at scale are both our people, and I want to be plain that this is not the same emergency for both. It is a slow leak for one and a direct hit for the other.

## Recipe blogs

A lot of the blogs we look after are food blogs, so this is where I watched it land first and hardest.

Recipes are the perfect thing for an answer engine to eat. How long do I rest this dough. What do I use if there is no buttermilk. Two line answers buried inside two thousand word posts, and a model lifts them cleanly.

They are also some of the heaviest pages on the web. Years of sensible decisions stack up: the recipe card plugin, the ad slots, the gallery, the video, six hundred comments. Googlebot is patient with that. It renders your JavaScript and comes back next week.

Agents are less patient. Google's own [guide to agent-friendly websites](https://web.dev/articles/ai-agent-site-ux) says agents read a site three ways, through screenshots, raw HTML and the accessibility tree, and calls the screenshot route "slow and expensive". Clean structure underneath the page stops being a nicety.

**Food bloggers got the worst of the trade. The content is exactly what these systems want, and the page is exactly what they struggle to read.**

## Two things you can check today

**1. Open the Generative AI performance report in Search Console.** For a long time the honest answer to "am I being cited" was that nobody could tell you. Google now [reports it](https://support.google.com/webmasters/answer/16984139) for its own AI features. Go and look before you buy anything from anyone.

**2. Decide about the AI crawlers on purpose.** These are separate from Googlebot, each documented by the company running it, and robots.txt controls them one by one. Almost nobody made that choice deliberately.

Here is the bit that trips people up, and I had it muddled myself for longer than I would like. **The crawler that trains a model and the crawler that gets you cited are not the same crawler.** Blocking the wrong one costs you the citations you were trying to win.

| Crawler | Run by | What allowing it does |
| ------- | ------ | --------------------- |
| OAI-SearchBot | [OpenAI](https://platform.openai.com/docs/bots) | Lets you show up in ChatGPT search answers |
| GPTBot | [OpenAI](https://platform.openai.com/docs/bots) | Lets your content train the foundation models |
| Claude-SearchBot | [Anthropic](https://support.anthropic.com/en/articles/8896518-does-anthropic-crawl-data-from-the-web-and-how-can-site-owners-block-the-crawler) | Lets your content be indexed for Claude search results |
| ClaudeBot | [Anthropic](https://support.anthropic.com/en/articles/8896518-does-anthropic-crawl-data-from-the-web-and-how-can-site-owners-block-the-crawler) | Lets your content train the models |
| PerplexityBot | [Perplexity](https://docs.perplexity.ai/guides/bots) | Lets you surface and get linked in Perplexity results |

There is a third group underneath those, the fetchers that run when a person actually asks a question: ChatGPT-User, Claude-User, Perplexity-User. Read their docs before you assume robots.txt covers them. OpenAI writes that because the action is user initiated, robots.txt rules "may not apply", and Perplexity writes that Perplexity-User "generally ignores robots.txt rules".

Not choosing is still choosing. If you want the citations, you have to let in the crawlers that actually produce them.

## What I will not sell you

This site runs the full kit. An [llms.txt](https://hostmy.blog/llms.txt) index, a clean Markdown copy of every page (add .md to this URL and see), and a WebMCP manifest at [/.well-known/mcp.json](https://hostmy.blog/.well-known/mcp.json) so an agent can ask structured questions instead of scraping. I would rather live with a thing for months than recommend it on a hunch.

**And Google, in [its own guide to generative AI search](https://developers.google.com/search/docs/fundamentals/ai-optimization-guide), files llms.txt under things "you can ignore for Google Search".** Their reason, in their words: "Google Search itself doesn't use them." The same page says structured data is not required for AI features, and that chopping posts into small chunks for the machines is not a thing you need to do.

So be precise about what these files are for. They are for the agents that do read them, and that group is growing: the [llms.txt proposal](https://llmstxt.org/) notes OpenAI, Anthropic and Gemini all publish one for their own docs. They are not a Google ranking tactic, and anyone selling them as one has not read the documentation.

What Google does say helps is duller and harder to fake. Crawlable pages, and content that is not a commodity. Their example of commodity content is a post called "7 Tips for First-Time Homebuyers". Their example of the opposite is a first-hand account of a decision somebody actually made.

**Which means the least automatable thing you own, your own experience, is the thing that now matters most.**

I find that reassuring rather than annoying.

## Where we are taking this

Here is the weakness in what I have described so far. Everything above is a thing you have to go and check by hand, on a Sunday, on a site with four hundred posts.

That is not a plan. That is homework.

So the direction we are building in is simple to say and slow to do. **We want to take the checking away from you entirely, and hand you the decision instead.**

Three pieces of that, in the order they matter to us.

- **Technical audits that run themselves.** Not a PDF somebody generates when you complain. A standing check of the things that decide whether a machine can read your work, running on your site, without you asking.
- **Reporting at the post level.** Site-wide scores are comfortable and mostly useless. The question worth answering is which post, which page, which fixable thing, and what it is costing you.
- **The work done, not just reported.** A finding you cannot act on is a worry with a chart attached. Where we can safely close the gap for you, we would rather close it.

Underneath all three is the same intent. **Getting a blog AI-ready is currently a months-long project for someone who already knows what they are doing. Compressing that is the entire reason Host My Blog exists in its new form.**

I want to be careful here, because this is exactly the point where most companies start promising. So, plainly: these are directions, not features you can switch on today. I am not giving you dates. When a piece of it is real, you will get it, and I will tell you what it does and what it does not.

The ambition behind it is not modest, though. Staying online is the floor. **We want to be the reason a blog does well, not just the reason it stays up.**

## RankReady, and one more thing

We packaged the plumbing into [RankReady](https://hostmy.blog/plugins/rankready/), free on WordPress.org. It writes the llms.txt, handles author identity, adds the schema, and gives you a switch per AI crawler so the decision above takes a minute. Free, because the alternative was repeating the same checklist to a different person every week.

There is another thing in the workshop. A way to run WordPress headless without the usual pain. It is called Hatch. It is not ready, and I will not promise you a date.

If you want the thinking rather than the plugin, [The Cited](https://hostmy.blog/the-cited/) is where I write it down. A couple of emails a month on what is genuinely getting sites quoted, including the attempts that failed.

## If you are already with us

You get all of it. The new look, the new experience, and much more of our attention on how AI search reads your site.

Your plan does not change. There is nothing for you to do. Your named person is still your named person.

Over the coming weeks the AI side of your blog gets the same care the speed and security side always had, and we will tell you what changed and why.

## Why we do this

Host My Blog is run by Etica Studio, in Ahmedabad. It was first called etica, Spanish for ethics, which sounds grander than what we actually do, which is answer our email and keep other people's sites online.

We started with one blog. We look after hundreds now, and every one arrived one at a time.

The part I am most protective of is small. A named person from our team looks after your site. The same person every time, not a ticket picked up by whoever is free. Someone who replies in hours rather than days.

The purpose has not moved. A blogger should be free to write. The server, the backups, the updates, the certificate, the speed, the security, all ours to carry.

The list just got one item longer, and the new item is whether a machine can read your work well enough to quote it.

**That is the whole idea. You keep the part only you can do, which is having something to say. We take the rest, including the part that is now machine-shaped.**

That is what the new site is built around, and I am glad it is out. Have a look. If something is wrong, write to me. I read every one.

Raj