What Unique Images Actually Cost Across 50, 200 and 500 Location Pages

There is a specific moment in a location-page project where the maths stops being abstract. You have the page template, you have the copy variables, you have forty towns in a spreadsheet, and then someone asks what image goes at the top of each one.

Nobody plans for this line item. It is usually the thing that turns a two-week project into a two-month one, not because images are expensive but because the cost is shaped differently from everything else in the build.

Why one image per page is the baseline

The tempting shortcut is a single hero reused across every page in the set. It costs nothing extra, it ships today, and it works in the narrow sense that no page is broken.

What it gives up is everything an image can contribute. One file is one asset: one filename, one alt attribute, one caption, one set of embedded metadata. Reused across forty pages it cannot describe forty different places, and the page-to-image relationship that makes an image findable in the first place is diluted forty ways. If several of those pages are otherwise near-identical, a shared hero makes them look more alike rather than less, which is the opposite of what a templated set needs. That mechanic is worth understanding on its own terms, because reused and near-identical images cause a specific class of indexing problem that is easy to create and slow to undo.

There is also the reader. A visitor who arrives on your Tulsa page and sees the same stock office interior they saw on the Wichita page has learned something true about how the pages were made.

So the working assumption is one relevant image per page, which makes the cost question real.

The three ways to get them

Generation produces a new file from a written description. Cost is a monthly allowance rather than a per-asset price, so the unit cost falls as you use more of the allowance and the marginal image is effectively free until you hit the cap.

Stock libraries licence existing photographs, priced per download or per credit pack. Cost is genuinely per image, so it scales linearly with page count, and the same catalogue is available to everyone including whoever else is building location pages in your vertical.

A photographer produces images nothing else can, priced per shoot. This is the right answer for places you physically occupy and a nonsensical one for the eighty towns you merely drive to, because there is nothing at those addresses to photograph.

Most real projects use two of the three: photographs of the actual premises and team, generation for everything that describes a service area rather than a location.

Running the numbers

Here is the shape of it at four project sizes, assuming one image per page and using SEOpix's published plan prices for the generation column. Handling time is separate and covered below.

Project size	Images needed	Generation plan	Effective per image	Stock	Photographer
50 pages	50	Starter, $19/mo (100 images)	$0.38	Scales with count	Impractical, no site to shoot
200 pages	200	Pro, $49/mo (500 images)	$0.25	Scales with count	Impractical
500 pages	500	Pro, $49/mo (500 images)	$0.10	Scales with count	Impractical
2,000 pages	2,000	Agency, $99/mo (2,000 images)	$0.05	Scales with count	Impractical

Two things fall out of that table.

The first is the direction of the curve. Generation gets cheaper per image the more you need, because the denominator grows while the numerator is fixed. Per-asset pricing does the opposite: the two-hundredth image costs what the first one cost. Somewhere below a hundred images the two are comparable and the decision is about taste and licensing rather than money. Somewhere above a few hundred, the comparison stops being close.

The second is that the pixel cost at any of these sizes is small. Five hundred images for $49 is not the number that decides whether the project happens. Which brings us to the number that does.

The cost that is not on the invoice

Every image needs the same handling regardless of where it came from: a descriptive filename, alt text, a sensible compression pass, any metadata you intend to embed, and an upload into the right place in the CMS.

Done by hand that is a few minutes per image if you are fast and know exactly what you are doing. Multiply by five hundred. At two minutes each that is roughly sixteen hours of someone's attention, and it is the tedious, low-judgement kind that degrades badly after the first hour. The images at the end of a manual batch are reliably worse than the ones at the start, which is how a location set ends up with forty careful alt texts and four hundred and sixty copies of the word "image".

This is the real argument for treating a bulk image set as a pipeline rather than a task. When the filename rule, the alt-text pattern and the metadata fields are decided once and applied by a process, handling stops scaling with volume. The mechanics of running a bulk image batch end to end are worth setting up before the first image exists rather than after the hundredth.

It is also why tools that write the metadata at creation time change the economics more than their sticker price suggests. The $49 buys pixels. The saving is in never opening five hundred files to type into them.

Where the money should actually go

A sensible budget for a location-page set looks less like a photography budget and more like this:

• Real premises, real team, real vehicles: photograph them. These cannot be substituted and are what a visitor most wants to see. They belong on your main pages and on location pages for places you genuinely occupy.
• Service-area pages: generate. There is no storefront in the town you drive to. An image that depicts the work honestly is more accurate than a stock photograph of somebody else's building.
• A reject budget. Assume ten to twenty percent of any batch will not be good enough. Size the allowance around the number you need to ship plus that margin, not the number of pages.
• A quality gate. Cheap images are only cheap if the bad ones are caught before publication. A short, repeatable pre-publish check on every image in a batch is what keeps the per-image cost honest, because an image redone twice cost three times what you budgeted.

The uniqueness question

The reason per-image cost matters at all is that the alternative to paying it is reuse, and reuse is where location pages go wrong.

This is not only about the hero. If every page carries the same photograph, the same alt-text pattern and the same caption, the images become one more thing telling an indexer that these pages are variations of a single page. Genuinely distinct imagery is among the cheaper ways to make a templated set feel like a set of real pages, which is worth more than the images themselves.

For local pages the image can also carry a location claim of its own. Embedded GPS coordinates say where an image belongs more precisely than a city name in a filename, and unlike a filename token it is a claim about the file rather than a keyword in a string. Whether that moves rankings on its own is unproven and anyone telling you otherwise is guessing. What is certain is that it costs nothing extra when written at generation time.

The licensing line nobody prices

There is a cost in the stock column that does not appear on the invoice, which is the ongoing obligation attached to the licence.

Standard stock licences carry conditions: limits on print runs, restrictions on use in logos or merchandise, prohibitions on implying endorsement by depicted people, and in some cases a requirement to stop using the image if the licence lapses. For one hero on one page that is trivial to track. Across five hundred location pages licensed over three years by four different people, it becomes a genuine record-keeping problem, and the failure mode is discovering you cannot prove your right to an image already published in five hundred places.

Generated images have a different shape of risk rather than none. The licence comes from the platform and varies between vendors, and copyright ownership of AI-generated work is treated inconsistently across jurisdictions, which matters if an image becomes a core brand asset rather than page decoration. What commercial use of AI-generated images actually permits is worth reading once, properly, before a large batch rather than after.

The practical guidance is the same either way: record what you licensed or generated, under what terms, and when. A spreadsheet with the file, the source, the date and the licence reference costs an hour to set up and answers a question that is otherwise unanswerable two years later.

What to actually do

Under about fifty images, pick on taste and licensing rather than price. The difference is a rounding error and any of the three routes is defensible.

Between fifty and a few hundred, handling already costs more than pixels, so choose whichever option lets you produce named, described, metadata-carrying files without a human touching each one.

Above a few hundred, per-asset pricing stops making sense and the only question worth arguing about is quality control. Get the rejection rate right and the rest of the economics follow.

If the images are for service pages rather than location pages the same logic applies with a different subject list, and a practical set of image ideas for service businesses is a better starting point than a stock search, for the same reason a generated workshop beats a licensed one: it can be about your work rather than about work in general.

Agencies running this across several clients hit one extra wrinkle, which is that each client needs their own consistent look rather than a shared house style. Keeping client image libraries separate and coherent is a workflow problem rather than a cost one, but it decides whether the economics above survive contact with five accounts.

As of September 2026, SEOpix prices generation at $0 for 10 images a month, $19 for 100, $49 for 500 with batches of 100 at a time, and $99 for 2,000 with sub-accounts, and writes the filename, metadata and geo-tags as each image is created. The full breakdown is on the pricing page, or you can start on the free plan and run ten images through your own pipeline before committing to a number.