Can an AI assistant read an image or PDF size chart?

That depends on the assistant, its access and the tools enabled in the session. Inspect what it actually received. A chart absent from a structured product response does not prove that every assistant is unable to read the image or document.

Does a measurement matching a chart row guarantee fit?

No. It establishes a match to that chart under its stated conditions. Keep the chart's measurement method, product scope and fit guidance visible, and ask for missing information rather than inventing a recommendation.

Size charts AI shoppers can use: verify the chart, units and answer

A shopper gives a measurement and asks which size to choose. Before evaluating the answer, establish which measurement the chart expects. Body dimensions, garment dimensions and internal product dimensions are different references, even when they use the same unit.

Start with the chart, not a recommendation: confirm the product link, expected measurement and unit. If those are unclear, resolve them before judging the assistant. Use the answer log for the session record.

Check the chart the answer is meant to follow

Pick one product and identify its applicable chart. Record the source, version, measurement method, unit and any conditions supplied by the manufacturer or your product team. Confirm whether a variant, product generation or market uses a different chart.

Do not infer the applicable chart from a nearby image or a shared “size guide” label. The relationship between chart and product is part of the evidence.

For a first test, choose a measurement away from a boundary. Then deliberately test the boundaries and incomplete questions. This separates a basic lookup problem from ambiguity in the guidance itself.

Inspect access without assuming the format failed

Scroll sideways for columns out of view. Keyboard: focus the table and use the Left and Right arrow keys.

Chart presentation	Useful check
Image or PDF	Did this assistant receive the file, and can you inspect its reading of the labels and values?
Page table	Were the relevant rows and headings present in the content it retrieved?
Pop-up or sizing widget	Was the chart opened or otherwise retrieved in that session?
Metafield or referenced record	Which representation exposed it, and was the correct product linked?

These are diagnostic questions, not a ranking of formats every agent can read. A missing explanation in the answer does not prove the rows were never read. Without retrieval evidence, record the mechanism as unknown.

Shopify supports connecting metafield content to a theme. Preserve a useful chart for human shoppers while checking its representation in the route you care about. For AI-channel product presentation, inspect Catalog Mapping; do not assume every chart automatically becomes a dedicated sizing tool.

Make the interpretation explicit

Here is a deliberately simple fictional chart for a garment. It is a data-review example, not sizing advice for a real product or a StoreKnows schema.

Scroll sideways for columns out of view. Keyboard: focus the table and use the Left and Right arrow keys.

Size	Body chest measurement	Unit	Boundary rule
S	88 to less than 94	cm	Includes 88; excludes 94
M	94 to less than 100	cm	Includes 94; excludes 100
L	100 to 106	cm	Includes both endpoints

Under this example’s stated rules, 98 cm maps to M and 94 cm also maps to M. A measurement of 107 cm is outside the chart. The chart does not support choosing the nearest size unless its owner supplies that rule.

Your real chart may use inclusive ranges that overlap, several measurements or qualitative fit preferences. Keep those rules. Do not quietly change a manufacturer’s boundaries to make the data easier to process.

For a maintainable representation, record the size label, measurement name, unit, limits and boundary meaning, with the chart’s product link and source date. Verify any chosen tool’s accepted formats before restructuring the data. JSON, a metaobject or a particular field name is not automatically a supported StoreKnows integration.

Review five answers before judging the intervention

Test an interior value, an exact boundary, an out-of-range value, a missing unit and a question using the wrong measurement type. Ask the responsible catalog operator what each should produce.

Keep the exact answer beside that reference. “Please confirm whether that is your body chest measurement in centimetres” may be the correct response. A confident size recommendation from incomplete information is not a better result.

Distinguish matching a chart from predicting personal fit; neither a structured row nor a simulated answer demonstrates fewer returns. For an intervention, use the comparison method with these five boundary cases.

Treat current variant availability separately from size guidance. A saved copy of sizing data is not a live stock check.

Decide whether a proposal is useful

Start with the native answer and your maintained chart. If that route already handles the important questions, extra publication may add no value. Verify a tool’s accepted chart formats before restructuring your data; the fictional garment chart above does not establish StoreKnows support for that chart. Saving or excluding a selection does not withdraw an older published bundle; verify the public version separately.

Run the free catalog-built check. If a supported size proposal appears, try your own question before deciding what to publish. How it works explains preview, publication and saved-data scope.

Corrected 13 September 2026: removed claims that all agents are unable to use images, PDFs, widgets or metafields, and removed unmeasured returns and automatic-withdrawal claims. The chart is illustrative, not a reported product test.

StoreKnows is independently developed and not affiliated with, endorsed by or sponsored by Shopify, OpenAI, Google or Anthropic.