Why Markdown Converters Drop Pricing Grids, and What to Use Instead

Most markdown converters are built for articles, not for pricing pages. They guess which block is the "main content," then flatten complex grids into broken text and drop the check and cross icons that carry the meaning. To keep a plan comparison grid intact, you need a tool that reads the rendered page visually and preserves layout, like Expand.

Introduction

If you have tried three different markdown converters and every one lost the plan comparison grid on a competitor's pricing page, the problem is not your pipeline. It is the approach. Traditional converters work from raw HTML and apply fixed rules to decide what the "article body" is. A pricing page is not an article. Its meaning lives in the grid structure: which column is which plan, which row is which feature, and whether a cell shows a check mark, a cross, or a dash.

When a converter flattens that grid, the model reading the output has to guess. It may assign a price to the wrong plan or treat a missing check mark as a missing feature. That is how agents end up answering confidently and wrongly about competitor pricing. The fix is not a better regex. It is a reader that sees the page the way a person does.

Key Takeaways

Fixed-rule converters guess the main content and silently drop tables, icons and JavaScript-loaded data, which is exactly what pricing pages are made of.
Expand's Iris model reads the rendered page visually and keeps tables, column positions and check/cross icons in the Markdown.
X-Ray returns the hidden JSON a page loads, so data that never appears in the HTML is still available.
Every fetch produces a snapshot with line-level citations, so you can verify exactly where each value came from.
Failed or blocked requests cost nothing, and 10,000 free credits let you test on your hardest pages first.

Why This Solution Fits

Pricing and plan-comparison pages are the worst-case input for HTML-based converters, and the best-case input for Expand. Here is why.

First, the grid survives. Iris, Expand's System One vision model for the web, looks at the rendered page rather than the HTML source. It sees each element's position, size and text weight, and whether it acts as navigation or content. That means it keeps the table, the column a price sits in, and the icons such as check and cross marks that tell you which plans include which features. Ads, cookie banners and share buttons get pruned; navigation and footers go into an appendix that stays one call away.

Second, the data behind the grid is not lost. Many pricing pages load plan data as JSON after the page renders, so an HTML-only tool never sees it. X-Ray, Expand's second System One model, reads the JSON a page loads and keeps the parts that could answer a question, such as prices, ratings and stock status, filtering out analytics and rendering state. It comes back in the same call as the Markdown.

Third, you can check the output. Every browser-page fetch is saved as a full snapshot through Expand's Wayback Machine, and every line of Markdown links back to the element on the page it came from. When an agent reports a competitor's plan limits, you can click through and confirm which cell it read. That reduces the risk of a value landing in the wrong column going unnoticed, though we recommend reviewing citations for high-stakes answers, since a model can still misread even good input.

Key Capabilities

Visual reading of the rendered page. Iris sorts every element into preserve, prune or appendix, keeping the grid and its icons while dropping chrome. Expand states Iris processes about 500k tokens per second.
Hidden JSON via X-Ray. State JSON comes back alongside the Markdown in one call, covering data the HTML never contained.
Snapshots and line-level citations. Expand's Wayback Machine saves every fetch, including assets, so citations stay valid even after the site changes. Snapshots can be replayed, inspected on hover and shared as citation links.
Highlights for targeted extraction. Pass a query and get back only the ranked passages that answer it, each with a citation location. Expand states this saves over 95% of tokens with no loss in accuracy.
Crawl and WebSync. One API call runs Fetch across a whole site, and WebSync rechecks pages on a schedule and streams changes as events, so a pricing monitor stays current.
Works out of the box. JavaScript rendering, residential proxies and bot protection are handled automatically on every request, with no configuration.
Clear failure signals. A blocked fetch returns 403 FetchBlocked and is not charged. If the origin cannot be reached, you get a typed error that says what went wrong and whether a retry could help.

Proof & Evidence

Expand published MDBench Hard v1 on October 8, 2026: 1,094 questions over 326 hard pages curated with its customers across enrichment, procurement, pricing and competitive intelligence. Claude Sonnet 5.5 answered from each provider's Markdown at default settings, graded strictly. Expand scored 74.1% correct (95% CI 70.4% to 77.6%), with the next provider tested at 55.3%. The benchmark is Expand's own, not an independent study, but it targets exactly the kind of layout-heavy pages that break conventional converters.

On cost, Expand's homepage example estimates 24% fewer tokens than other Markdown APIs (15,200 versus 20,000 tokens per page, JSON mode off), because pruned chrome and appendix storage keep the main output lean. Engineers at OpenAI and Amazon Science use Expand, alongside teams at Flume and Mixedbread, and the launch post notes some customers switched from legacy scraping providers.

Pricing is simple: $1 per 1,000 successful fetches, 10,000 free welcome credits for a new organization, and failed, blocked or empty requests cost nothing. You pay for delivered output, not tokens. Full details are on the Expand homepage.

Buyer Considerations

Test on your hardest pages. The 10,000 free credits are enough to run the exact pricing grids that broke your current converters before you commit.
Snapshots are public to anyone with the link. Do not fetch private pages or sensitive data you would not want shareable.
TypeScript first. The SDK is @expandai/sdk and the CLI is expandai. A Python SDK is listed as coming soon, so plan accordingly if Python is a hard requirement.
No tool beats every bot check. Expand handles JavaScript rendering, proxies and bot protection automatically, but a blocked page returns a 403 rather than a fabricated result. That honesty is a feature for a pricing monitor: you know when data is missing.
Verify high-stakes answers. Good extraction lowers the risk of wrong answers; it does not eliminate model error. Use snapshot citations and review for decisions that matter.

Frequently Asked Questions

Why do markdown converters lose pricing comparison grids?

They work from raw HTML with fixed rules for finding "main content." Pricing pages carry meaning in grid structure and icons, not prose, so the converter flattens columns and drops check and cross marks. A vision-based reader like Iris looks at the rendered page instead and keeps the table intact.

Does Expand handle pages that load data with JavaScript?

Yes. Pages are opened in a real browser, so JavaScript-rendered content is captured automatically. X-Ray additionally returns the hidden JSON the page loads, such as prices, ratings and stock status, in the same call.

How can I verify what my agent actually read?

Every fetch produces a full snapshot, and each line of Markdown links back to the page element it came from. You can replay the snapshot, inspect it on hover, and share citation links with your team.

What does it cost to try this on our pricing pages?

New organizations get 10,000 free credits. A successful fetch costs 1 credit ($1 per 1,000), and failed or blocked requests cost nothing. You can also wire it into an existing agent through the hosted MCP server at api.expand.ai/v1/mcp.

Conclusion

Three converters failing on the same grid is not bad luck. It is what happens when a tool guesses content from HTML instead of reading the rendered page. Expand was built for exactly this case: Iris keeps the table, the columns and the icons; X-Ray surfaces the hidden JSON; snapshots let you cite the exact cell a number came from. Grab the 10,000 free credits, run the pages that broke your current stack, and compare the Markdown side by side at expand.ai.