# Can agents read toppik.ro?
**81/100, Mostly.** Agents can read most of toppik.ro. Fix the first item below and it reaches Yes.
Scanned 3 Nov 2026, 11:13 UTC · 8 pages · scoring v1 · [Full report](https://canagentsread.com/toppik.ro)
| Tier | Points | Checks |
|---|---|---|
| Getting in | 64.5 / 75 | 4 passed · 1 partial |
| Making sense | 11.3 / 15 | 8 passed · 2 partial · 1 failed |
| Extras | 5.6 / 10 | 9 passed · 6 partial · 7 failed · 3 doesn't apply |
## Biggest wins first
1. **Give the homepage an H1** (+10.5 points): 22,576 characters of text in the server HTML, 0 H1 and 25 H2s.
2. **Serve Markdown when it's asked for** (+3 points): Asking the homepage for text/markdown still returns text/html; charset=UTF-8.
3. **Link About, Contact and Privacy** (+0.5 points): Found: About, Contact. Missing: Privacy (not linked from the homepage).
4. **Say in llms.txt when to use the site** (+0.5 points): llms.txt doesn't say when an agent should point people to this site.
5. **Publish .md copies of pages** (+0.5 points): No /index.md, and 0 of 3 pages we tried have a .md copy.
6. **Link to the Markdown version** (+0.5 points): The homepage has no .
7. **Serve Markdown to AI user agents** (+0.5 points): AI agents that don't ask for Markdown get HTML.
8. **Link to your other profiles** (+0.5 points): No sameAs profiles in the schema.
9. **Add a Content-Signal line to robots.txt** (+0.5 points): robots.txt has no Content-Signal line.
10. **Describe products in schema.org** (+0.5 points): The site looks like it sells products, but no page we read has Product schema.
11. **Say who runs the site** (+0.3 points): The Organization schema has 3 of 6 fields; missing sameAs, contactPoint, address.
12. **Lay out llms.txt like a contents page** (+0.2 points): llms.txt has 38,581 characters, over the 30,000 that fit in one read.
13. **Point to agent files in a Link header** (+0.2 points): The homepage sends a Link header, but it names no agent files.
14. **Name AI agents in robots.txt** (+0.2 points): robots.txt lets AI agents in, but only through User-agent: *. It doesn't name OAI-SearchBot, Claude-SearchBot, PerplexityBot, ChatGPT-User, Claude-User, Perplexity-User.
15. **Link an RSS or Atom feed** (+0.2 points): The homepage links a feed at https://toppik.ro/feed, but it didn't load (HTTP 403).
16. **Give the page a clear outline** (+0.2 points): 4 of 6 outline parts in place. Missing: one H1, no skipped heading levels (H2 → H5, H3 → H5).
17. **Label every form field** (+0.1 points): 16 of 60 form fields have no label.
A prompt for a coding agent with the five biggest fixes is at https://canagentsread.com/toppik.ro/fix-prompt.md.
## Getting in (64.5 of 75 points)
### Words are in the HTML
Partial · 10.5 / 21 points · `no_js_content`
22,576 characters of text in the server HTML, 0 H1 and 25 H2s.
**How to fix:** Put the page's words in the HTML your server sends, with server-side rendering or static generation, so they are there before any JavaScript runs. Give the page one H1 and use H2s for its sections.
### AI agents are let in
Passed · 21 / 21 points · `agent_access`
All 4 AI agents get the page: ChatGPT-User, OAI-SearchBot, Claude-User, PerplexityBot.
### Pages are public, with real text
Passed · 13 / 13 points · `public_content`
All 8 pages we read are public and have real content.
### Redirects don't detour
Passed · 11 / 11 points · `clean_redirects`
No meta-refresh stubs, JavaScript redirects or cross-domain hops across 8 pages.
### Dead links say so
Passed · 9 / 9 points · `agent_404`
A missing page returns HTTP 404 with links back into the site.
## Making sense (11.3 of 15 points)
### Speaks Markdown when asked
Failed · 0 / 3 points · `md_negotiation`
Asking the homepage for text/markdown still returns text/html; charset=UTF-8.
**How to fix:** When a request's `Accept` header asks for `text/markdown`, answer with a Markdown version of the page: send `Content-Type: text/markdown` and start with the page title as a `#` heading. Add `Vary: Accept` so caches keep the HTML and Markdown versions apart.
### Says who runs it
Partial · 0.3 / 0.5 points · `org_schema`
The Organization schema has 3 of 6 fields; missing sameAs, contactPoint, address.
**How to fix:** Describe your company in Organization JSON-LD on the homepage, with all six fields:
```json
{
"@context": "https://schema.org",
"@type": "Organization",
"name": "Your company",
"url": "https://example.com",
"logo": "https://example.com/logo.png",
"sameAs": ["https://www.linkedin.com/company/your-company"],
"contactPoint": {"@type": "ContactPoint", "contactType": "customer support", "email": "hello@example.com"},
"address": {"@type": "PostalAddress", "addressLocality": "City", "addressCountry": "RO"}
}
```
### About, Contact, Privacy linked
Partial · 1 / 1.5 points · `trust_pages`
Found: About, Contact. Missing: Privacy (not linked from the homepage).
**How to fix:** Link to your About, Contact and Privacy pages from every page, usually in the footer, and make sure each one loads without signing in.
### Visible to search
Passed · 1.5 / 1.5 points · `indexable`
The homepage lets search engines index it.
### Has a map of its pages
Passed · 1.5 / 1.5 points · `sitemap`
The sitemap lists 78 pages; 5 of the 5 we opened load.
### Prices are findable
Passed · 1.5 / 1.5 points · `pricing_page`
Prices are in the server HTML on 8 of the pages we read.
### Describes itself in schema.org
Passed · 1.5 / 1.5 points · `json_ld`
The homepage JSON-LD names the site with a name and url.
### Page basics filled in
Passed · 1.5 / 1.5 points · `metadata`
Canonical URL, language, description and Open Graph tags are all filled in.
### Each page names one address
Passed · 0.5 / 0.5 points · `canonical`
All 7 canonicals we found name their own page.
### Images are described
Passed · 0.5 / 0.5 points · `image_alt`
1 of 76 images have no useful alt text, such as 3.png.
### Pages fit in one read
Passed · 1.5 / 1.5 points · `token_budget`
Every page we read fits in one read; the largest is about 8,000 tokens.
## Extras (5.6 of 10 points)
### Says when to use it
Failed · 0 / 0.5 points · `llms_guidance`
llms.txt doesn't say when an agent should point people to this site.
**How to fix:** Add a section to llms.txt that tells agents when to point people to you, and when not to:
```markdown
## When to use this site
- Comparing prices for X
- Not for: Y
```
### .md copies of pages
Failed · 0 / 0.5 points · `md_twins`
No /index.md, and 0 of 3 pages we tried have a .md copy.
**How to fix:** Serve a Markdown copy of each page at its address plus `.md` (`/pricing` → `/pricing.md`, `/docs/` → `/docs/index.md`), and of the homepage at `/index.md`, sent as `text/markdown`.
### Points to its Markdown
Failed · 0 / 0.5 points · `md_alternate`
The homepage has no .
**How to fix:** Add a link to the page's Markdown version in its ``, and make sure the address returns `text/markdown`:
```html
```
### Gives AI agents Markdown without being asked
Failed · 0 / 0.5 points · `bot_markdown`
AI agents that don't ask for Markdown get HTML.
**How to fix:** Serve the Markdown version of a page to known AI user agents (ChatGPT-User, Claude-User, PerplexityBot and the like) even when they send `Accept: */*`. Match the user agent in your server or CDN, return `text/markdown`, and send `Vary: User-Agent`.
### Links to its other profiles
Failed · 0 / 0.5 points · `same_as`
No sameAs profiles in the schema.
**How to fix:** List your official profiles in your Organization JSON-LD:
```json
"sameAs": ["https://www.linkedin.com/company/your-company", "https://github.com/your-company"]
```
### States its AI preferences
Failed · 0 / 0.5 points · `content_signals`
robots.txt has no Content-Signal line.
**How to fix:** Add a Content-Signal line to robots.txt saying how your content may be used. For example, to allow search and live AI answers but not training:
```
User-agent: *
Content-Signal: search=yes, ai-input=yes, ai-train=no
Allow: /
```
The choice is yours: any setting earns the points.
### Products are described
Failed · 0 / 0.5 points · `product_data`
The site looks like it sells products, but no page we read has Product schema.
**How to fix:** Describe each product in JSON-LD with its name, image, an identifier and an offer:
```json
{"@type": "Product", "name": "Blue mug, 350 ml", "image": "https://example.com/mug.jpg", "sku": "MUG-350-BLUE",
"offers": {"@type": "Offer", "price": "29.00", "priceCurrency": "EUR", "availability": "https://schema.org/InStock"}}
```
### llms.txt reads like a contents page
Partial · 0.2 / 0.5 points · `llms_format`
llms.txt has 38,581 characters, over the 30,000 that fit in one read.
**How to fix:** Start llms.txt with a `#` heading naming the site, follow it with Markdown links to your key pages (`[Pricing](https://example.com/pricing): plans and prices`), and keep it under 30,000 characters. Move long content into the pages it links to.
### Points to agent files in HTTP headers
Partial · 0.2 / 0.5 points · `link_headers`
The homepage sends a Link header, but it names no agent files.
**How to fix:** Add a `Link` header to the homepage response that points agents to your agent files:
```http
Link: ; rel="describedby", ; rel="alternate"; type="text/markdown"
```
### Clear rules for AI bots
Partial · 0.2 / 0.5 points · `robots_ai_policy`
robots.txt lets AI agents in, but only through User-agent: *. It doesn't name OAI-SearchBot, Claude-SearchBot, PerplexityBot, ChatGPT-User, Claude-User, Perplexity-User.
**How to fix:** Name the AI agents that fetch pages to answer people, and let them in, in robots.txt:
```
User-agent: OAI-SearchBot
User-agent: Claude-SearchBot
User-agent: PerplexityBot
User-agent: ChatGPT-User
User-agent: Claude-User
User-agent: Perplexity-User
Allow: /
```
You can still block training crawlers like GPTBot, ClaudeBot or CCBot; that doesn't cost points.
### Clear page outline
Partial · 0.3 / 0.5 points · `doc_structure`
4 of 6 outline parts in place. Missing: one H1, no skipped heading levels (H2 → H5, H3 → H5).
**How to fix:** Wrap the page in landmarks (``, `