To get a WordPress site cited by ChatGPT and Perplexity, let their crawlers in, give them plain text that states who you are and what you charge, and write pages that answer a question in the first paragraph. On topsyde.com that means an llms.txt file, a Markdown pricing page, an explicit robots.txt and answer-first posts.
What does it take to get cited by ChatGPT and Perplexity?
An answer engine can only cite a page it was allowed to fetch, could read, and found a quotable passage in. Those are three separate gates. A site can fail any one of them without anyone on the team noticing, because the site still loads fine in a browser and still ranks in Google.
The business case is plain arithmetic about where buyers ask questions now. According to OpenAI, ChatGPT reached 800 million weekly active users (2025). According to Pew Research Center, Google users who saw an AI summary clicked a traditional search result on 8% of visits, against 15% on visits with no summary, and clicked a link inside the summary itself on 1% of visits (2025). When fewer people click through, the business named in the answer gets the shortlist spot, and the ten blue links below it get less attention.
When an agency owner asks ChatGPT "which managed WordPress host includes developer time," we want topsyde.com in the answer. So we set the site up for that, and below is the exact setup, with the parts you can copy this week.
What is llms.txt, and what does ours say?
llms.txt is a plain-text Markdown file at the root of a site that summarizes what the business does and links to its most useful pages. Jeremy Howard of Answer.AI proposed it in September 2024. Ours says what TopSyde does, who it is for and what it costs, in language a model can lift without guessing.
It is a proposal. No major AI company has confirmed it reads llms.txt when choosing what to cite, and Google's John Mueller said in 2025 that no AI service had said it uses the file, comparing it to the old keywords meta tag. We publish one anyway, for two reasons. It costs an hour to write. And when someone pastes our URL into Claude or ChatGPT and asks "what does this company charge," the model fetching that page has a clean, current summary available instead of a JavaScript pricing toggle.
The format is short. A small business version looks like this:
# Harbor Plumbing
> Licensed residential plumber serving Portland, Oregon since 2011.
> Emergency calls answered 7am to 10pm. Drain clearing from $149.
## Services
- [Drain clearing](https://harborplumbing.com/drains): pricing and what is included
- [Water heaters](https://harborplumbing.com/water-heaters): repair and replacement
## Pricing
- [Price list](https://harborplumbing.com/prices.md): every service and fee
Write it the way you'd describe the business to a new hire on their first morning: who you serve, where, what it costs, and where the details live. If you'd rather have Claude act on your site than read it, that is a different tool, and our guide to connecting WordPress to Claude through an MCP server covers it.
Why publish pricing as a Markdown file?
Pricing pages are where answer engines most often get a business wrong, because prices sit inside toggles, tabs and scripts that a crawler may not run. A Markdown file puts every plan in one plain table. Ours, /pricing.md, lists every TopSyde plan side by side with no script required to read it.
The content matches our pricing page exactly, as closed ranges:
| Plan | Sites | Price per site, per month |
|---|---|---|
| Standard | 1 to 10 | $89 |
| Growth | 11 to 99 | $71 |
| Enterprise | 100 to 199 | $67 |
| Agency | 200 to 299 | $57 |
| Agency Plus | 300 and up | $49 |
Source: TopSyde plan prices as published 1 November 2026. Yearly billing saves 10%, registered nonprofits save 10% on monthly plans, and the two don't combine.
Closed ranges matter here. If your page says "from $49," a model can repeat "$49" to someone with three sites. Writing "$71 a site a month for 11 to 99 sites" leaves nothing to misread. We think every service business with public prices should ship a file like this, and update it the same day the prices change.
Which AI crawlers should robots.txt allow?
Allow the crawlers that feed products which name and link their sources, and decide case by case on crawlers that only collect training data. Our robots.txt explicitly allows seven AI-related user agents and blocks one. Here is each one, who runs it, and what we set.
| User agent | Run by | What it does | topsyde.com setting |
|---|---|---|---|
| GPTBot | OpenAI | Collects public pages that may be used to train OpenAI models | Allowed |
| ChatGPT-User | OpenAI | Fetches a page when a ChatGPT user asks it to | Allowed |
| PerplexityBot | Perplexity | Indexes pages so Perplexity can surface and link them in answers | Allowed |
| ClaudeBot | Anthropic | Collects public pages for Anthropic's models | Allowed |
| anthropic-ai | Anthropic | Older Anthropic user agent, listed for coverage | Allowed |
| Google-Extended | A robots.txt token that controls whether Google may use your pages for Gemini | Allowed | |
| Bingbot | Microsoft | Crawls for Bing, whose index also feeds Copilot | Allowed |
| CCBot | Common Crawl | Builds an open web archive that many AI labs train on | Blocked |
Source: topsyde.com robots.txt, checked 1 November 2026.
Two details catch people out. First, Google-Extended does not decide whether you appear in Google's AI Overviews; those draw on the regular Googlebot crawl. Blocking Google-Extended keeps your pages out of Gemini training and grounding, and does nothing to Search. Second, AI companies now split crawlers by job, and new names keep appearing: OpenAI runs OAI-SearchBot for ChatGPT search, and Anthropic documented Claude-SearchBot and Claude-User in 2025. If your robots.txt blocks everything not on a list, check it lets those through.
A minimal version for a WordPress site:
User-agent: GPTBot
Allow: /
User-agent: ChatGPT-User
Allow: /
User-agent: PerplexityBot
Allow: /
User-agent: ClaudeBot
Allow: /
User-agent: CCBot
Disallow: /
Sitemap: https://example.com/sitemap_index.xml
Robots.txt is only a request. If your CDN or firewall turns the bot away first, the file never gets read. Cloudflare began blocking known AI crawlers by default for new domains in July 2025, so a site behind Cloudflare can have a welcoming robots.txt and still be closed to every AI crawler. Check the bot settings in your CDN dashboard as well as the file.
Why does topsyde.com block CCBot?
We block CCBot because it is a training-only crawler with no citation benefit. Common Crawl builds a free archive of the web that many AI labs train on, but no product built on that archive links back to us or names us as a source. Allowing it gives our pages away and earns nothing measurable in return.
That is a judgment call, and it is the opposite of what we do with GPTBot and ClaudeBot, which are also training crawlers. The difference is the company on the other end. OpenAI and Anthropic run products that browse, search and cite. If their models already know TopSyde exists and what it costs, a buyer asking a broad question is more likely to hear our name. Common Crawl has no such product. If your goal is to appear in as many models as possible, you might allow CCBot; if your goal is citations with a link, blocking it costs you nothing we can find.
How should a page be structured to get quoted by AI?
Put a complete answer of 40-60 words directly under each heading, phrase headings the way people type questions, and end with an FAQ. Answer engines pull passages that stand on their own. A paragraph that only makes sense after the previous three rarely gets lifted.
Every post on this blog follows that template, and this one is no exception. The opening paragraph answers the title. The TLDR gives four to six bullets that each work as a quote with no context. Each H2 is a question someone would type, and the paragraph under it answers that question before expanding. The FAQ at the end adds more question and answer pairs, which is the format a chat interface returns.
None of this is exotic. It is the inverted pyramid newspapers have used for a century, applied to every section instead of only the top. Most of the work is cutting the warm-up paragraph that business blogs open with. If you're on TopSyde and the first paragraph of your homepage or service page buries the answer, AI edits included with hosting let you rewrite that text from a chat, preview it, and undo it for 30 days.
Why should comparison tables be sourced and dated?
A price without a date is a guess the moment it's published, and answer engines repeat numbers whether or not they're current. Every competitor figure on our /compare page is sourced and dated. That gives a model, and a reader, a way to check it, and it tells both how old the number is.
Here is how we format it, using care companies checked on their own pricing pages:
| Company | Plan | Per site, per month | Hosting | Checked |
|---|---|---|---|---|
| WP Buffs | Maintain | $89 | Extra below the Custom plan | 21 October 2026 |
| FatLab | Starter | $99 | Included | 21 October 2026 |
| GoWP | Business | $99 | Extra | 21 October 2026 |
| TopSyde | Standard, 1 to 10 sites | $89 | Included, plus 1 hour of developer time a month | Our price, 1 November 2026 |
On WP Buffs, malware removal and speed work start on the Perform plan at $239 a month. Writing that down with a date is what turns the table into something an engine can cite with confidence, and something a competitor can't call misleading. Copy the habit: a "Checked" column, or a source line under every table, and a calendar reminder to recheck quarterly.
How do WordPress homepages score on AEO/GEO?
In our sample, poorly. Between 3 September and 30 October 2026 we ran 515 WordPress homepages through our free site audit. The median AEO/GEO score was 55 out of 100, second lowest of the six categories, one point above structure. None of the 515 sites were TopSyde clients.
| Audit category | Median score out of 100 |
|---|---|
| SEO | 70 |
| UX | 66 |
| Overall | 58 |
| Performance | 56 |
| AEO/GEO | 55 |
| Structure | 54 |
Source: TopSyde site audit results for 515 WordPress homepages, 3 September to 30 October 2026.
Know the limits before you quote this. Each result is one homepage, one load, from one location. The sites are businesses and agencies we chose to contact, so the numbers describe that sample and not WordPress as a whole.
Even with those limits, the pattern is useful. The two lowest medians, AEO/GEO and structure, are the two categories closest to whether an answer engine can understand and quote a homepage. SEO scored highest, which fits: most of these sites have an SEO plugin doing the traditional work. What tends to be missing is the layer above it.
The AEO/GEO category asks whether a homepage gives an answer engine something it can read and quote: who the business is, what it does and for whom, in text a crawler can parse. You can run the same audit on your own homepage at topsyde.com/audit and compare your number against that median of 55. Page weight is a separate story (Elementor homepages in the sample had a median of 41 scripts against 32 for Gutenberg), and we covered it in the 515-homepage page builder benchmark and in our comparison of Full Site Editing and page builders.
What can a WordPress site owner copy this week?
Most of this setup takes an afternoon and costs nothing. Open your live robots.txt, check your CDN's bot settings, write llms.txt, publish prices in plain text, rewrite your homepage's first paragraph and date your comparison tables. None of it guarantees a citation; it removes the reasons an engine would skip you.
- Visit yourdomain.com/robots.txt to read your live file. WordPress serves a virtual one by default; Yoast SEO and Rank Math both include an editor, and a physical file uploaded to the site root replaces the virtual one.
- Look for an "AI bots" or "bot fight" setting in Cloudflare or your host's dashboard. A block there overrides anything robots.txt says.
- Write llms.txt using the format above and upload it to the root over SFTP, or let Yoast SEO or Rank Math generate a starting version and edit it by hand.
- Put prices in plain text: a pricing.md, or at minimum a plain HTML table, with closed ranges.
- Rewrite the first paragraph of your homepage and top service pages so it states who you serve, what you do and roughly what it costs in 40-60 words.
- Add a source and a checked date to any table with someone else's numbers in it.
- Ask the engines what they already say. We did this for TopSyde and wrote up what ChatGPT said about our business; run the same prompts for yours.
One more thing worth checking. Crawlers read whatever your public HTML contains, including spam links injected by a compromised plugin. If your pages carry text you didn't write, AI answers can repeat it. Free online scanners only see that public layer, and we explain what free WordPress malware scanners can't see on the server side.
What does TopSyde include for $89 a month?
TopSyde is managed WordPress hosting at $89/mo per site for 1 to 10 sites, with $1,800+ a year of developer time and premium plugins included. That is 1 hour of developer time a month per site, worth $840 a year at a typical $70 freelance rate, plus premium plugin licences worth $1,000+ a year at list price.
For the AI search work described above, that included hour is enough to cover a robots.txt review, an llms.txt and a pricing file on a typical small business site, and the 10 AI edits per site per month cover rewriting opening paragraphs without opening a ticket. Developer work beyond the included hour is $150 an hour or quoted first. Agencies running 11 to 99 sites pay $71 a site a month; the agency hosting page covers the rest, and the spec sheet lists what every plan includes.
New monthly plans start with a 30-day free trial. You add a card up front, nothing is charged that day, and the first charge comes when the 30 days end unless you cancel first. Migration is free and plans run month to month. If you're coming from a budget host, our write-up on moving from GoDaddy to managed WordPress hosting shows what the move involves. Start with the free audit, see your AEO/GEO number, and decide from there.
Frequently Asked Questions
Does llms.txt help a site get cited by ChatGPT?
No AI company has confirmed it uses llms.txt when choosing sources, and Google's John Mueller said in 2025 that no AI service had said it reads the file. We publish one because it takes an hour and gives any model fetching our site a clean summary of what we do and charge. Treat it as cheap insurance, and put the real effort into robots.txt access and answer-first pages.
Should I block GPTBot to protect my content?
That depends on whether you want OpenAI's models to know your business exists. GPTBot collects pages for training, while ChatGPT-User fetches pages live when someone asks, so blocking GPTBot alone doesn't remove you from browsing answers. We allow GPTBot because a model that already knows our name and pricing is more likely to mention us.
Will blocking Google-Extended remove my site from AI Overviews?
No. Google-Extended only controls whether Google may use your pages for Gemini training and grounding. AI Overviews draw on the regular Googlebot crawl, so blocking Google-Extended leaves your Search and AI Overview eligibility unchanged.
What is a good AEO/GEO score in the TopSyde audit?
The median in our sample of 515 WordPress homepages, audited between 3 September and 30 October 2026, was 55 out of 100. That sample is businesses and agencies we chose to contact, measured on one homepage load each, so use it as a reference point rather than a pass mark. Run the free audit and see where your homepage sits against it.
Can I set this up on Wix or Squarespace?
Partly. Hosted builders control their own robots.txt rules and limit which files you can place at the site root, so check what your platform allows before planning around llms.txt. Our comparison of what a five-page site costs on Wix, Squarespace and WordPress covers what you can and can't change on each.
Founder & Lead Developer
20+ years full-stack development, WordPress, AI tools & agents
Colton is the founder of TopSyde with 20+ years of full-stack development experience spanning WordPress, cloud infrastructure, and AI-powered tooling. He specializes in performance optimization, server architecture, and building AI agents for automated site management.
