Crawler access
Blocked or missing pages.
What this means: Crawler access
An AI crawler is blocked, gets a 404, or never reaches a page that matters.
Chapter 01 of 05
Autonomous AI SEO for existing websites.
Each night, the engine finds supported SEO issues, publishes fixes within your rules and checks the live HTML. Changes are delivered without editing the client’s CMS.
- Find SEO issues
- Apply your rules
- Check core changes
88
automated health checks
13
check areas
25
fix points deployed at the edge
30 min
between live-page rechecks
The loop
Find. Fix. Deliver. Verify.
Supported fixes for existing pages, delivered within your rules and checked after publication. No new articles.
One issue → a checked changeILLUSTRATIVE WORKFLOW
01
Find
88 automated health checks read the site and set the health score.
02
Fix
The change is written, inside the guardrails you set.
03
Deploy
Served in the HTML at the edge, following your publishing rules.
04
Verify
A real browser re-opens the page after title, meta, H1 and schema deploys.
05
Roll back
Automatically when the core post-deployment checks detect a regression.
Your publishing rules apply at every step
Your control level determines what can publish. Risky fixes get a safety review; you control visible rewrites per site. Every step is logged.
Step 1 of 5 · Find
Identify technical SEO issues. 88 automated health checks.
Review site health, prioritize findings and enable supported fixes. Each delivered change is recorded in the activity log.
What an issue can look like Illustrative examples
Structured data
Missing structured page information.
What this means: Structured data
A page a model would cite has no schema to cite from.
Page title and meta
Missing or unclear search descriptions.
What this means: Page title and meta
Title or description is missing, empty, or does not say what the page answers.
Canonical and duplicates
Two addresses. No clear original.
What this means: Canonical and duplicates
Two addresses show the same thing, and neither is marked as the original.
Entity signals
Business details that do not match.
What this means: Entity signals
Name, address and phone number differ between pages, so the model is unsure who the client is.
Freshness
Pages that need another look.
What this means: Freshness
A page that used to be cited has sat untouched and is losing ground.
Internal linking
Pages with no path leading to them.
What this means: Internal linking
A page no other page points to is a page the model rarely finds.
Indexing signals
Discovery instructions that disagree.
What this means: Indexing signals
Sitemap, robots and IndexNow disagree about what exists.
The technical SEO checklist
Explore the 13 check areas. Scheduled crawls process pages in batches.
The technical SEO checklist
Explore the 13 check areas. Scheduled crawls process pages in batches.
88health checks
Know what needs attention.
Checks cover 13 areas, from page titles to crawler access.
Finding an issue does not mean it can be fixed automatically.Page titles
5 checksFind missing, duplicate or poorly sized titles.
Meta descriptions
4 checksFind missing, duplicate or poorly sized descriptions.
Headings
7 checksCheck the main heading and heading structure.
Content
10 checksFind thin content, duplicate pages and heavy resources.
Links
9 checksFind broken links and pages that are hard to reach.
Images
2 checksCheck alternative text and image file sizes.
URLs
3 checksFind inconsistent or overly long page addresses.
Canonical links
5 checksCheck which version of a page should be indexed.
Structured data
4 checksCheck schema markup and missing fields.
Sharing tags
11 checksCheck the tags used for social previews.
Indexability
12 checksFind settings or errors that prevent discovery.
Hreflang
8 checksCheck language and regional page relationships.
Domain level
8 checksCheck HTTPS, discovery files and crawler access.
See the full checklist All 13 areas explained
- Page titles · 5 checks
- Missing, too short, too long, duplicated across pages, more than one on the page.
- Meta descriptions · 4 checks
- Missing, too short, too long, duplicated across pages.
- Headings · 7 checks
- Missing or multiple main headings, skipped levels, heading identical to the title, repeated across pages, and length.
- Content · 10 checks
- Thin, duplicate, low quality, oversized page or HTML, missing language, broken resources, heavy scripts and stylesheets, insecure mixed content.
- Links · 9 checks
- No internal links, too many links, empty link text, empty addresses, broken internal and external links, orphan pages, invalid attributes, pages with only one incoming link.
- Images · 2 checks
- Missing alt text and oversized image files.
- URLs · 3 checks
- Double slashes, uppercase letters, overlong paths.
- Canonical links · 5 checks
- Missing, invalid, pointing at another page, pointing outside the domain, or more than one.
- Structured data · 4 checks
- No schema on the page, invalid schema, missing recommended fields, and schema Google itself rejected.
- Sharing tags · 11 checks
- Open Graph and X/Twitter: missing blocks, missing image, missing or invalid card type, and text that gets cut in the preview.
- Indexability · 12 checks
- Error pages, redirects, noindex in the page or the header, gaps in the sitemap, non-indexable pages listed in the sitemap, conflicting robots directives, nofollow, meta refresh, insecure redirect targets.
- Hreflang · 8 checks
- Missing return links, no default, missing self-reference, invalid language codes, ambiguous regional codes, duplicates, broken targets, lone alternates.
- Domain level · 8 checks
- HTTPS, serving of sitemap and robots, unclassified site type, expiring certificate, and AI crawlers blocked for search, training or mixed purposes.
Additional technical coverage: 24 extra checks from an independent second engine, plus an external crawl for duplicates, redirect loops and non-indexable pages.
Audit views & prioritization
Inspect findings by page and priority, with an assistant that can explain the site’s data.
Audit views & prioritization
Inspect findings by page and priority, with an assistant that can explain the site’s data.
Understand site health.
See the whole-site score, what passes and what needs attention.
Find pages that get missed.
Explore the site tree. Spot orphan pages and content buried too deep.
Know what to work on next.
Review recommendations in priority order, with the site’s data behind each finding.
Explore the audit views and controls
- Health score 0–1000
- One score for the whole site, with the week’s change and what moved it. The list also shows what the site already passes, not just what it fails.
- Structure tree and page explorer
- The whole site as a navigable hierarchy, plus every page in a sortable table with score, status and focus keyword.
- Orphan pages and crawl depth
- Pages nothing links to, and pages buried so deep they are invisible in practice. Internal link suggestions are found by meaning, not word match.
- Content pruning report
- Every page sorted into keep, thin or prune — judged against actual Google impressions instead of a guess. Reversible with one click.
- Readability and brand voice
- Readability score, sentence length and how closely a page hits the voice you defined. Advisory only — never published to the live page.
- Ignore a finding
- Hide a check for one page or the whole site without skewing the score. Plus an independent second audit with 24 extra technical checks.
- Recommendation feed
- A prioritized list of what to do next, scored daily by likely impact — so the most valuable item sits on top instead of the newest.
- Publishing pace
- How often the client publishes, and how many new pages arrive each month. Makes changes in the site’s publishing rhythm visible over time.
- Site type detection
- The engine works out what kind of site the client runs — store, blog, local business, publisher — and picks the structured data that fits that type.
Help with your client’s data
Ask about the site.
Know where to start.
Your questionWhere should I start?
SEO assistant
Review high-priority findings, then open the affected pages.
Page findingsRead-only. It explains the data and links you to the relevant screen. It cannot publish changes.
Understand the findings.
Answers use the client’s actual site data.
Plan the work.
Ordered steps link to the screen where you work.
See what runs next.
Review upcoming automation and the reason behind it.
Find the right screen.
Search pages, sites and settings from the keyboard.
Explore the assistant and navigation tools
- SEO assistant
- A conversation that knows the site’s actual numbers and answers with concrete, linked steps.
- Action plans
- Ask for a plan and get ordered steps, each linking straight to the screen where the work is done.
- Queue for what comes next
- What the automation intends to work on next, and why, before it runs.
- Command palette
- Jump to any page, site or setting from a keyboard search.
Step 2 of 5 · Fix
What it fixes. Item by item, in the HTML at the edge.
Browse common fixes below, or explore the complete capability catalog.
Catalog scope
Explore The complete open list of 358 capabilities across 24 areas in thecapability catalog: 328 capabilities in 23 areas are client-facing. 30 capabilities in 1 area are agency and white-label functions.
62items in total
- 25 deployed at the edge
- 28 measured or reported
- 9 control
Technical foundation
What search engines and models read first. Most of it is fixed in the HTML that is served, without anyone changing how the client publishes; the rest is checked and reported.
Open an item to see what it covers.
Deployed at the edgeTitle tags and meta descriptions
Write unique titles and descriptions for the subject each page covers. Add them where they are missing.
Deployed at the edgeImage alt text
Give meaningful images descriptive alt text that explains their subject.
Measured or reportedCore Web Vitals
Loading (LCP), interactivity (INP) and visual stability (CLS), measured from real Chrome visitors and reported per page.
Deployed at the edgePreload, lazy loading and modern image formats
Preload hints and critical CSS for the first screen, lazy loading below the fold, and picture markup for AVIF or WebP — applied where the client’s host can serve them.
Measured or reportedMobile friendliness
Google reads the mobile version first. PageSpeed audits check and report mobile problems; the viewport setting is corrected where the page already carries one.
Deployed at the edgeSitemap.xml and robots.txt
The map and the rulebook the engines get: where the pages are, and where not to look.
Deployed at the edgeCanonical tags
Says which page is the original, so variants of the same page are not read as duplicate content.
Deployed at the edgeSchema.org / structured data
The language machines speak. Price, availability and ratings are marked up only when the page actually shows them.
Deployed at the edgeBreadcrumb markup
The breadcrumb trail is marked up, so the engine can show the path to the page instead of just the URL.
Deployed at the edgeHeading structure and H1
One main heading per page and logical subheadings. You control visible rewrites per site.
Deployed at the edgeRedirects
The 301 and 302 redirects you create are live at the edge immediately, and a server that redirects to its own internal address is sent back to the client’s domain.
Deployed at the edgeRobots directives
What gets indexed and what does not, with a filter so only valid instructions go out. Clearing a stray noindex goes through a safety review first.
Deployed at the edgeLanguage attribute
Every page declares its correct language, validated, and matching the version actually served.
Measured or reportedHTTPS
The certificate for the connected site is issued and renewed automatically, and an expiring certificate on the client’s own server is flagged before it takes the site down.
Measured or reportedHreflang checks
Eight checks find missing return links, invalid language codes and broken targets between language versions. The launch offering includes writing hreflang tags at the edge.
Deployed at the edgeIndexNow
Search engines that support IndexNow are told the moment a page changes, instead of waiting for the next crawl.
Deployed at the edgeOpen Graph and X/Twitter cards
How the page reads when shared on Facebook, LinkedIn and X: title, description and card type — added where missing. An existing share image is kept; an image address that might not exist is never invented.
Deployed at the edgeSecurity headers
Modern security headers are added to every response served, without touching the client’s server setup.
Deployed at the edgeHome-screen and app icons
Touch icon, tile color and theme color, so the site looks right when someone saves it to their phone.
Deployed at the edgeSEO plugin markup set aside
Rank Math, Yoast, All in One SEO and Squirrly are detected and their competing schema is replaced by one merged graph, so machines do not get two different answers.
Measured or reportedSchema coverage report
Which valuable schema types the site still lacks, page by page — out of 31 types the engine can write.
Measured or reportedURL hygiene
Double slashes, uppercase letters and overlong paths are flagged before they become an indexing problem.
Measured or reportedBroken links and resources
Internal links, outbound links, scripts, stylesheets and images are tested for errors and reported, so you decide what gets redirected.
Deployed at the edge410 for retired pages
Mark a page permanently gone, so search engines and AI crawlers drop it — instead of redirecting it forever.
Deployed at the edgeOrganization identity
Legal name, logo and official profile links marked up and tied to the Wikidata identifier, so machines can verify who the client actually is.
Keywords and pages
Which queries the client can actually win, and which page should answer each of them. Measured and reported; the findings feed the rewrites.
Open an item to see what it covers.
Measured or reportedKeyword research
No guessing. Search data shows which words customers use, including misspellings and local variants.
Measured or reportedPrimary keyword per page
One clear lead term per page, which the title and heading rewrites are written around.
Measured or reportedSemantic coverage
Related terms that make the topic unambiguous — write about Apple, and iPhone and iOS help the engine know which Apple.
Measured or reportedLong tail and question queries
Fewer searches per phrase, but visitors who know what they want.
Measured or reportedKeyword mapping to pages
Every keyword gets its own home, so five pages do not fight over the same term.
Measured or reportedGap analysis against competitors
The client site laid over a competitor’s to see where content is missing.
Measured or reportedCompetitor analysis
Which terms do they rank for that the client lacks? The holes in their strategy become the client’s work list.
Measured or reportedInternal anchor text
Link text tells the engine what the target page is about. “Read about running shoes” beats “click here”.
Measured or reportedOngoing SERP monitoring
The result page changes — map packs, videos, images, AI answers. The features are tracked, so you see when the picture changes.
Measured or reportedMovement report
Which terms bring the most traffic, and which are on the way up. The cold facts.
Content and structure
How the pages connect, and how the models understand what belongs to what.
Open an item to see what it covers.
Measured or reportedInternal link opportunities
Pages that should link to each other, found by meaning and reported. The related-links box can add them to the page once visible rewrites are on.
Deployed at the edgeFAQ and HowTo markup where it fits
Questions, answers and steps marked up only where the page actually contains them.
Measured or reportedDuplicate control
Pages that are too alike are found and reported; a canonical can mark the one that counts.
Deployed at the edgeCanonical for variants
The same product in several colors or sizes points to one original.
ControlAlerts on breakage
The smoke detector: if traffic drops or an important page disappears, you hear about it.
ControlNavigation is never touched
Visible rewrites deliberately skip everything in the menu, footer and sidebar.
Deployed at the edgeDecorative images are left alone
Purely decorative images are recognized and skipped — no description is written or forced.
Deployed at the edgeRight-to-left languages
Correct text direction for Arabic, Hebrew and similar languages, also in text the engine inserts.
Deployed at the edgeMap box from coordinates
Map boxes for physical addresses are built safely from coordinates instead of pasted third-party code.
Local visibility and authority
The signals that tie the client’s name to a place and a field — and to other sources that mention them.
Open an item to see what it covers.
Deployed at the edgeStructured NAP
Name, address and phone marked up the same way on every page, so the engine is never in doubt about who the client is.
Deployed at the edgeKnowledge Graph link
The client’s Wikidata identifier tied to the pages, so the engine connects the site to the entity it already knows.
Measured or reportedLink analysis
Who links to the client? Weak links are surfaced and new opportunities found.
Measured or reportedDirectory presence
Whether the client is listed in the leading local directories in its market — the listings that confirm the business is real.
Measurement
Reporting against GA4 and Search Console, and on who the models cite — with the numbers that actually moved.
Open an item to see what it covers.
Measured or reportedMonthly report
The report sums up what was changed and what moved, per client — white-label, ready to forward.
Measured or reportedGA4 and Search Console
Measured against the client’s own accounts, not invented numbers. Everything reported can be checked.
Measured or reportedCitations in AI answers
Scheduled prompts are sent to the live engines with web search, in the client’s market, and the answer is read: who is cited.
Measured or reportedWhich AI crawlers visited
Read straight from the request at the edge, not from a tracking script — a crawler runs none.
Measured or reportedImpressions, clicks and keywords
How often the client shows up, how often someone clicks, and on which queries.
Measured or reportedMovement per page
Every page is followed on its own, so you see which ones lift and which ones stand still.
Measured or reportedWhat changed, and what worked
The report separates changes that were made from changes that actually moved a number.
Control and insight
The engine works on its own; control stays with you the whole way, in real time.
Open an item to see what it covers.
ControlFull log of every change
Every change listed with its time, the page it applied to and what was changed.
ControlSee what is live, any time
The panel shows the current state at all times — not yesterday’s report.
ControlWithdraw a change with one click
Disagree? Withdraw it yourself. The previous state is served again on the next request.
ControlAdjust a change yourself
Want to keep the change but phrase it differently? Edit it in the panel.
ControlSwitch to review when you want
Want to look before anything goes live? Set the client to Review everything, and every change waits for your approval.
ControlNo plugin in the CMS
The source code in the publishing system is never modified. Everything is delivered in the HTML response at the edge.
ControlNo visible widget for visitors
The work happens in the markup the engines read. The design visitors see stays exactly as the client wants it.
Autopilot
Choose what publishes automatically.
Set a control level per client. It determines which enabled changes can publish automatically and which require your team’s approval.
Level 1Default
Full autopilot
Enabled changes publish within your rules.
- Search signals
- Automatic
- Technical fixes
- Automatic
- Visible text
- Automatic when enabled
- Risky fixes
- After safety review
Level 2
Invisible-only
Automate technical work. Review visible text.
- Search signals
- Automatic
- Technical fixes
- Automatic
- Visible text
- Your approval
- Risky fixes
- After safety review
Level 3
Review everything
Your team approves each change before it goes live.
- Search signals
- Your approval
- Technical fixes
- Your approval
- Visible text
- Your approval
- Risky fixes
- Your approval
Your rules stay in place.
Choose a level per site. Enable visible rewrites separately. Menus, footers and sidebars stay untouched.
See checks and rollbackVisible text publishes automatically only on Full autopilot, when enabled and within your rules.
Risky fixes need a safety review before automatic publication and are flagged for your team.
What do these changes include?
- Search signals
- Title, meta description and structured data
- Technical fixes
- Missing canonical, language attribute and social tags
- Visible text
- H1, subheadings, link text, captions, answer blocks
- Risky fixes
- Clearing a stray noindex or a conflicting canonical
After publication
The work keeps running.
Automatic work. Clear next steps for you.
Find the next work.
Crawls revisit existing pages and discover new ones. Unfinished batches continue in later runs.
Keep the work in place.
Compare the live page with the expected version. Restore changes that were overwritten.
Deliver panel changes.
Changes made in your panel reach visitors without republishing the original site.
Review the fixes that need you.
Actions outside automatic work wait for your review.
Start with the highest priority.
Daily recommendations are ranked by likely impact.
How the ongoing work runs
- Nightly run
- Scheduled crawls revisit the site and discover new pages. Work is processed in batches, with unfinished pages continued in later runs.
- Self-repair every 30 minutes
- The live page is compared with what it should be. If something overwrote the work, it is put back.
- One-click fixes waiting for you
- What the engine does not do on its own is collected as clear buttons you can review and press when you are ready.
- Recommendation feed
- Next steps are scored daily by likely impact, so the most valuable item sits on top instead of the newest.
- Live in about a minute
- A change you make in the panel reaches real visitors within about a minute. No republish, no queue.
Step 2 of 5 · Fix, one example
What a change looks like.
A fictional bike-service page, before and after supported changes to its HTML. The client’s original site stays in place.
Northwind Cycles · bike service pageSample
The same page. Clearer signals.
Before → afterOriginal pageDelivered HTML
Page title
Home | Northwind Cycles
Bike Service in Seattle — Book Today
Description
Empty
Service for every bike. Reply within one day.
Image description
Missing
Bike service in the workshop
Changes are delivered in the HTML response. The client’s CMS stays unchanged.
See the HTML changes6 sample changes
Original HTML
<title>Home | Northwind Cycles</title><meta name="description" content=""><img src="workshop.jpg">Delivered HTML · changes at the edge
<title>Bike Service in Seattle — Book Today</title><meta name="description" content="Service for every bike. Reply within one day."><link rel="canonical" href="https://northwindcycles.com/bike-service"><meta property="og:title" content="Bike Service in Seattle — Book Today"><img src="workshop.jpg" alt="Bike service in the workshop"><script type="application/ld+json">{"@type":"Service"}</script>Step 3 of 5 · Deploy
Preferium delivers the optimized page.
Our changes are already in the HTML when it arrives. Here’s why we chose this approach over browser-only fixes and crawler-specific delivery.
WHY WE DELIVER AT THE EDGE
A cleaner path from your site to the reader.
Our approachServer-side delivery
Preferium AI Edge
The optimized page arrives ready to read.
We apply changes before sending the HTML. On server-rendered sites, visitors and crawlers receive the same optimized page.
- No SEO script for the reader to run
- No crawler matching to receive the changes
- Your original CMS content stays untouched
Competitor methods & their trade-offs
OTTO · browser pixel
Changes depend on JavaScript.
LimitationIf a crawler does not run the script, it misses those changes and reads the original HTML.
OTTO also documents an edge mode.Compare OTTO delivery
Alli AI · crawler rendering
The crawler must be recognised.
LimitationA separate rendered page is selected for detected crawlers. Delivery depends on that detection step.
Alli also offers a DNS integration.Compare Alli AI delivery
For JavaScript-only sites: Preferium also needs dynamic rendering to make the page readable to crawlers.
Delivery sources & comparison scope
These cards compare delivery methods, not every mode each vendor offers. OTTO's limitation above applies to its browser pixel, not its documented edge mode. Alli's card describes its documented crawler-rendering approach. It is not a measured failure rate.
Vendor documentation checked 3 October 2026: OTTO rendering modes, Alli crawler rendering and Alli DNS integration.
Steps 4 and 5 of 5 · Verify and roll back
Checked on the live page, withdrawn if it got worse.
A real browser checks core page elements after deployment. See what triggers an automatic rollback and what needs your attention.
Re-open the live page.
A real browser returns after deployment, once the edge serves the new version.
Check the core elements.
Compare title, meta description, H1 and structured data. Missing or empty elements can trigger rollback.
Undo a worsening change.
The change is rolled back automatically. An alert tells you what was withdrawn.
The checks also flag a lost page body or most of its content. Other edge changes remain available for manual rollback.
All verification and rollback rules
- What is re-checked
- After every deploy, a real browser re-opens the live page and re-checks title, meta description, H1 and structured data.
- When
- A minute or two after the deploy, once the edge is serving the new version.
- What counts as worse
- The page lost its body or most of its content, or the deployed element is empty or missing on the live page.
- What happens then
- The change is rolled back automatically, and you get an alert saying what was withdrawn.
- Drift, every 30 minutes
- The live site is compared with what it should be. Work something overwrote is repaired, and you are told if it could not be restored.
- Cloaking guard
- Each optimized page is fetched as a person and as GPTBot. A serious difference is flagged to you.
- Everything else
- Stays reversible with one click: undo per change, withdraw per element, and the last 50 automated changes listed with undo.
Answer blocks
Add concise on-page answers when visible rewrites are enabled for the site.
Answer blocks
Add concise on-page answers when visible rewrites are enabled for the site.
On the client’s pageFictional example
Northwind Cycles
Short summary
Bike servicing in Seattle.
Workshop servicing for your bike. Contact the shop to choose a time.
Direct answer
How do I book a service?
Tell the shop which bike you ride and what needs attention. Agree on a service and a time before bringing it in.
Key takeaways
- Describe your bike and the issue.
- Choose a service with the shop.
- Confirm a time to bring it in.
SourcesWorkshop services · Booking information
Answer blocks
A clear answer, inside the existing page.
Summaries, direct answers and key points help readers find what they need.
- Short summary
- Direct answer paragraph
- Key takeaways
- Sources
Visible rewrites must be enabled. Your publishing rules still apply.
Labels follow the page’s language, in 21 languages, including right-to-left text for Arabic and Hebrew.
What else can change on the page?
Related links box
Up to seven internal links chosen by meaning, not word match — helps both readers and crawlers find related pages.
Map box from coordinates
For physical addresses, built safely from coordinates instead of pasted third-party code.
List items and table text
Bullet points are rewritten to be complete, and tables get a caption and column headers machines can understand.
Code explanations and captions
A short plain-language explanation above code examples, and captions that say what the image actually shows.
Form labels and link text
Clear labels and helpful placeholder text — and “click here” becomes text that says where the link goes.
Duplicate titles avoided while writing
The titles of nearby pages are shown to the AI as it writes, so duplicates are avoided immediately instead of being found in an audit weeks later.
Inside an answer block
- Short summary
- One or two sentences that say what the page answers — the first thing a model reads.
- Direct answer paragraph
- A paragraph written as an actual answer, not as sales copy.
- Key takeaways
- Three to five points an assistant can lift verbatim without rewriting them.
- Sources
- Where the claims come from, so the answer can be checked — by the model too.
Files and discovery signals
See what Preferium generates and which AI-use choices belong to your agency.
Files and discovery signals
See what Preferium generates and which AI-use choices belong to your agency.
Generated by Preferium
A clearer way into your site.
Files and signals help crawlers find pages and understand the site.
- /sitemap.xml
- /robots.txt
- /llms.txt
- /agent.json
- /agents.txt
- /feed.xml · Atom
- Web app manifest
- /.well-known/security.txt
- IndexNow
Discovery support, not a guarantee of indexing or AI citations.
Chosen by your agency
You set the AI-use signal.
Declare how AI may use the content.
- Search
- AI answers
- Training
Your declaration communicates your choice. It does not force a crawler to comply.
Explore each file and signal Purpose and responsibility
| File or signal | What it does | Handled by |
|---|---|---|
| /sitemap.xml | Every page with sensible priorities; blocked and non-indexable pages are left out. | Preferium |
| /robots.txt | AI crawlers are allowed by default, and the client’s own rules are respected. | Preferium |
| /llms.txt | A site overview following the llms.txt proposal — in short and full versions. | Preferium |
| /agent.json | Machine-readable manifest describing the site for AI agents. | Preferium |
| /agents.txt | The same description in text form, for agents that look for it. | Preferium |
| Content signal | A clear declaration of what AI may do: appear in search, answer questions, train — your choice. | Your agency |
| /feed.xml · Atom | Content feed generated from the client’s pages. | Preferium |
| Web app manifest | So the site behaves like an app when saved to a phone. | Preferium |
| /.well-known/security.txt | Published security contact, so researchers can reach the client the right way. | Preferium |
| IndexNow | Search engines are notified the moment a page changes, instead of waiting for the next crawl. | Preferium |
Your agency chooses the content signal. Preferium generates the other discovery outputs. Sitemap priorities and update frequency can be set per page; thin or private pages are excluded.
Loading, delivery & languages
Measure real visitor performance, review supported loading fixes and connect language versions.
Loading, delivery & languages
Measure real visitor performance, review supported loading fixes and connect language versions.
Measure the visitor experience.
Core Web Vitals use real Chrome visits. PageSpeed audits inspect performance and page quality.
Load the important files first.
Supported fixes prioritize the first screen, use modern image formats and defer images below it.
Connect the language versions.
Check language links and regional codes. Use a glossary to protect words that must stay unchanged.
Use a safe copy when available.
If an origin fetch fails or times out, an eligible cached copy can serve public pages. A usable copy must already exist.
Performance and delivery details
- Core Web Vitals
- Loading, responsiveness and visual stability, taken from real Chrome visitors.
- PageSpeed audits
- Google’s own audit of performance, SEO, accessibility and best practices — automatic and on demand.
- Critical CSS and preload
- The styles for the first screen are inlined, and the files that matter most load first.
- AVIF and WebP
- Picture markup so the browser can fetch a much smaller version, with a safe fallback to the original.
- Lazy loading
- Images and embeds below the fold load only when the visitor scrolls to them.
- Cached fallback for public pages
- If an origin fetch fails or times out, an eligible cached copy can serve public pages. This requires a usable copy already in the cache; authenticated and personalized requests are not served from this shared fallback.
Language and international SEO details
- Eight hreflang checks
- Finds missing return links, invalid language codes, ambiguous regional codes and broken targets.
- Translation with grammar
- Titles, meta and headings translated into the languages you want, with correct regional grammar.
- Hreflang tags and a glossary
- Hreflang tags link language versions, and a glossary protects words that must stay unchanged. Included in the launch offering.
- Serving per language
- Every visitor and crawler gets the right version, with a sensible fallback when there is no exact match.
- Inserted text in the page’s language
- Answer block, links box and map labels write their own labels in 21 languages.
Want to see it on a client site?
Pick one client site for a walkthrough of the engine and your branded panel.
Check the site
Review its health checks.
See the first fixes
Understand what the engine would improve.
Explore your panel
See the client experience under your brand.
Your client siteYour guardrailsClean HTMLYour brand
Preferium AI Edge
One connection. A clear way back.
- Connect a client site
- Set the control level
- Run under your brand