Install Retouchable in ChatGPT Retouch your product photos from a ChatGPT chat. Install in ChatGPT
  1. Turn on Developer mode

    In ChatGPT, open Settings → Security and login and switch on Developer mode.

  2. Copy the address, open ChatGPT

    https://api.retouchable.ai/api/mcp

    Copy & open ChatGPT ↗
  3. Create the app

    On ChatGPT's Plugins page, select + and choose Create MCP App. Name it Retouchable, paste the address, set authentication to OAuth and create it.

  4. Sign in to Retouchable

    ChatGPT opens our sign-in. New to Retouchable? That starts your 14-day free trial, no card.

Then, in a chat, choose Retouchable under Developer mode and ask: “Grade the photos on my store” or “Put my product photo on white.”

Needs ChatGPT Plus, Pro, Business, Enterprise or Edu, on the web. Menus look different? OpenAI's steps

On-Site Search Results: Thumbnails That Earn the Click

A shopper who types into your search box already knows what they want. The thumbnail's job is to prove, at a glance, which result is it.

|conversion optimization e-commerce image optimization Shopify

A shopper who uses your on-site search has done something a browser never does: they have told you, in words, what they want. "Black leather ankle boot." "Linen duvet king." "Gold hoop small." Every thumbnail on the search results page is then read against that sentence. The image is no longer asked to attract attention. It is asked to confirm a match, and a thumbnail that can't confirm the match loses the click even when the product is exactly right.

That makes search results a different problem from the collection grid. A collection page is a selection a merchandiser made, and the shopper is browsing it. A search results page is a set your search engine assembled in milliseconds, often with partial matches and near-misses mixed in, and the shopper is checking it. The photo is how they do the checking, because reading twelve product titles on a phone is slower than looking at twelve pictures.

This post is about that checking job: what a searcher looks for in a result thumbnail, how to make the image answer the query instead of the product page, and how to measure whether search thumbnails are costing you clicks. For where the featured image appears across a Shopify store, including predictive search and the size it renders at, see our map of where the Shopify featured image shows. We won't repeat that here.

Why on-site search thumbnails do a different job

Browsing and searching are different states of mind, and product images serve them differently. A browser on a collection page is open to suggestion: the image's job is to make one product more appealing than its neighbours. A searcher has a specification in their head and is filtering against it. For them, the most useful image is the one that lets them say yes or no fastest.

Three things follow from that.

  • Relevance beats appeal. A beautiful lifestyle shot of a boot on a cobbled street loses to a plain shot that clearly shows the boot is black, leather and ankle height, if those are the words in the search box.
  • Every result is compared with every other. Search results put similar products next to each other more often than any curated page does. Whatever separates them has to be visible at result size.
  • A miss costs more than a non-click. If the image misleads (shows the brown colourway to someone who searched "black"), the shopper either skips a correct result or opens a wrong one and comes back. Both read in your analytics as search that isn't working, when the search itself was fine.

Collection grid shopper

  • Browsing a set someone chose
  • Asks: "which of these do I like?"
  • Image wins by being appealing
  • Products usually shot as a set
  • Mood and context help

Search results shopper

  • Checking a set the engine assembled
  • Asks: "which of these is the thing I typed?"
  • Image wins by confirming the query
  • Products from different shoots side by side
  • Attributes they named must be visible

The thumbnail has to confirm the query

Most search queries are a noun plus one or two attributes: a colour, a material, a size, a style word. The attributes are what the shopper will check in the image, so they are what the image has to show. Go through your top search terms and ask, for each, whether the result thumbnails make the named attribute obvious without reading the title.

Attribute in the queryWhat the thumbnail must showCommon failure
Colour ("black", "sage")The colourway searched for, true to lifeFeatured image is a different variant
Material ("leather", "linen")Surface texture readable at small sizeSoft, flat lighting erases grain and weave
Shape or cut ("ankle", "wide-leg", "hoop")The full silhouette, front-on or three-quarterTight crop or angled shot hides the outline
Size ("small", "king", "mini")A scale cue: in hand, on a bed, on a modelIsolated packshot with nothing to measure by
Set or count ("set of 4", "pair")All the pieces in frameOne piece shown; the set is in image 3

Colour is the failure merchants see most, because the featured image of a product with variants is one colourway, usually the first one shot. Search for "green" and the result may be the right product with the wrong thumbnail. Some themes and search apps can swap in the image of the matching variant when a shopper filters by colour, and some can do it when the colour is in the query text. Others always show the product's featured image. Check yours: search your own store for a colour that is not your featured variant and see what appears. If the thumbnail doesn't change, every searcher for the other colours is looking at a picture of a product they didn't ask for. Clean, consistent variant images, the same ones we cover in variant swatch images, are what makes that swap possible when the theme supports it.

A quick check for each top query

Search the term on a phone. Cover the titles with your thumb. Can you point to the results that match every word you typed? If you hesitate on any of them, that thumbnail is failing a searcher.

Separating near-misses: the image as a relevance signal

No store's search ranks perfectly. A query for "ankle boot" returns ankle boots, but also Chelsea boots, knee boots with "ankle strap" in the description, and boot socks tagged "boots". Search engines match words; shoppers match things. When ranking is imperfect, the thumbnail becomes the relevance signal: it is how the shopper throws out wrong results without opening them.

That is good news for you. A shopper who can discard a near-miss in half a second keeps scanning. A shopper who can't tell whether a result is a match has to open it, find out it isn't, and come back, and many don't come back. Every result your images let a shopper dismiss quickly is time saved for finding the right one.

For that to work, products that are easily confused need images that make the difference visible:

  • Shoot related products the same way. An ankle boot and a Chelsea boot shot from the same angle on the same background are easy to tell apart, because the only difference in the frame is the product. Shot on different days in different setups, the backgrounds and angles differ too, and the shopper has to work out which differences matter.
  • Show the distinguishing feature from the front. If the difference between two products is elastic side panels or a zip, the featured angle should show it.
  • Keep accessories out of the frame. A sock shown inside a boot makes the result look like a boot. An accessory product's featured image should be the accessory alone.
  • Don't let a bundle masquerade. A gift set that contains a candle will appear in a search for "candle". If its thumbnail shows only the candle, it reads as a single candle at a set price.

None of this is new photography advice. What's different is the reason. A collection page rarely puts a Chelsea boot next to an ankle boot by accident; search does it constantly.

Search results are a comparison table

Once a shopper has narrowed the results to the ones that match, they compare. On a product page they'd read specifications. On a results page they compare what the thumbnails let them compare. For the attributes that make a shopper choose one match over another, the image is the spec sheet.

Which attributes matter depends on the category, and a store's own search terms usually say which. A rule of thumb: the deciding attribute should be visible at the size your search results render on a phone, and it should be shown the same way in every product in that category.

CategoryDeciding attributes at result sizeMake them comparable by
FootwearToe shape, heel height, soleSame side-profile angle for the whole range
BagsSize, strap type, closureSame distance and height; strap shown, not tucked
JewelleryScale, finish, stone colourSame crop scale; one worn shot in the set for size
Home textilesColour, texture, pattern scaleSame fold or drape; raking light for weave
FurnitureShape, leg style, materialSame three-quarter angle; same plate colour

The last column is the part most catalogues miss. Comparison only works if the images are taken on the same terms. If one sofa is shot at three-quarter and the next head-on, the shopper is comparing camera angles as much as sofas. Consistency is a conversion tool here, not a brand nicety, because it lets the shopper see the only thing that differs.

Watch the product scale across a result set

A small product shot to fill the frame will look the same size as a large product shot the same way. In search, where a mini bag and a weekend bag can sit side by side, that misleads. If size is a deciding attribute in a category, keep a consistent relationship between product size and frame size across that category, or include a scale cue in the image.

Measuring how search thumbnails perform

Search is unusually easy to measure because every session starts with a stated intent. You can see what someone wanted and whether they clicked a result. That makes it possible to find thumbnail problems query by query rather than guessing at them.

The figures worth looking at, whatever tool provides them:

  • Search click-through by query. Of the searches for a term that returned results, how many led to a click on one? A term with plenty of results and few clicks is the strongest signal that the results look wrong, whether or not they are.
  • Refinement rate. How often is a search immediately followed by another, more specific search? "Boots" then "black ankle boots" is normal narrowing. "Black ankle boots" then "black leather ankle boots" after a look at the results may mean the thumbnails didn't show leather.
  • Click position. If searchers consistently skip the first few results to click the sixth, look at the first few thumbnails. The ranking may be fine and the images unconvincing.
  • Quick returns. A click on a result followed by a fast return to the results page suggests the image promised something the product page didn't deliver: the wrong colour, the wrong size, the wrong item.

Where to find these depends on your setup. Shopify's analytics include reports on online store searches, and dedicated search apps typically report search terms, click-through and searches with no clicks in more detail. Google Analytics can track site search if it is configured to read your search URL's query parameter. Any of these is enough to rank your top queries by how badly their results convert to clicks.

Separate the two problems

A search with no results is a search problem: synonyms, spelling, missing products. A search with good results and no clicks is often an image problem. Don't fix the second with the tools for the first.

Once a query is flagged, change its thumbnails and compare the same metric before and after over comparable periods. Seasonal swings move search volume a lot, so compare click-through rate, not raw clicks. Our guide to A/B testing product images covers how to keep a before-and-after honest.

A search-thumbnail audit in an afternoon

You don't need a full catalogue review. Search traffic is concentrated: a store's top few dozen queries usually account for a large share of all its searches. Fix the thumbnails behind those and you have fixed most of what searchers see.

  1. Pull your top 25 search terms for the last 90 days, with click-through for each if your tool provides it.
  2. Run each search on a phone and screenshot the first screen of results. That first screen is what most searchers judge.
  3. Mark each result as a clear match, a clear non-match, or unclear. "Unclear" is the one that costs you: a shopper can't decide from the image.
  4. Name the cause of each unclear result using the attribute table above: wrong variant, attribute not visible, inconsistent angle, misleading scale, accessory or bundle confusion.
  5. Group the fixes. Wrong variants are often a theme or app setting, or a variant image that was never uploaded. Inconsistent angles are a reshoot or re-edit of a category. Missing scale cues mean adding or promoting an in-hand or on-model image.
  6. Re-run the searches after the fix and record the query metrics as a baseline for next month.
25Top queries to audit
1Screen of results per query
3Verdicts: match, non-match, unclear

The audit usually turns up a short list of products that appear across many top queries. A tote bag that shows up in searches for "tote", "canvas bag", "beach bag" and "work bag" is judged by its one thumbnail in all four. Those few products are where an image fix has the most reach.

Fixing result thumbnails without a reshoot

Most of the fixes the audit turns up are not about taking new photos. They are about which existing photo leads, and whether it was prepared for the job.

  • Promote a better image to position 1. In Shopify the first image in a product's media is the featured image, and that is what search results show by default. Often the right image is already in the gallery, in position three or four.
  • Make the range consistent. Products from different shoots can be brought onto the same background colour, a similar crop and consistent light, which is most of what makes a result set comparable. This is editing work, not photography.
  • Fill in the variant images. If your theme or search app can show the variant that matches the query, it can only do so when each variant has its own image.
  • Keep the product honest. A thumbnail edited to look better than the product leads to the quick returns that the metrics above flag. Edits should change the background, light and framing, not the product.

This is where Retouchable fits. Phone photos or older catalogue shots can be brought onto a consistent studio background and light in batches, and with the Shopify connection a finished image goes straight back to the product: you choose the filename, write the alt text and set its position, including position 1, so it becomes the image search results show. The connection only asks for permission to read and write products, nothing about orders or customers.

Write the alt text for people and for search engines, describing the product and the attribute a shopper would search for ("Black leather ankle boot with side zip"). Whether your storefront search reads alt text at all depends on the search engine behind it, so treat the title, product type, tags and variant names as what your on-site search matches on, and the image as what the shopper checks the match against. For the first image's wider role beyond search, see why the first image decides click-through.

Frequently Asked Questions

Do product images affect on-site search rankings?

Usually not directly. Most storefront search engines rank on text such as the product title, type, tags, vendor and variant names. Images affect what happens after ranking: whether the shopper recognises a result as a match and clicks it. A product that ranks first with an unclear thumbnail can lose the click to a lower result whose image confirms the query.

Why does my search show the wrong colour for a product?

Search results show the product's featured image by default, which is one colourway. Some themes and search apps can swap in the matching variant's image when a shopper filters by colour or names one in the query, but only if every variant has its own image. Check your theme and search app settings, and make sure each colour variant has an image assigned.

Should search results use lifestyle images or plain product shots?

For most categories, a plain, consistent product shot works better in search, because the shopper is checking attributes and comparing results side by side. Lifestyle images help when the deciding attribute is scale or fit, such as jewellery size or how a garment sits on the body, but they should still show the searched-for attribute clearly.

How do I know if my search thumbnails are costing me sales?

Look at click-through by search term. A query that returns plenty of relevant results but gets few clicks, or where shoppers routinely skip the top results or search again with more specific words, usually points to thumbnails that don't confirm the match. Run the search yourself on a phone to see what the shopper sees.

Make every search result look like the match it is

Bring your catalogue onto a consistent studio look in batches and push the image you choose to position 1 of each Shopify product.

Try Retouchable Free No credit card required