Your Restaurant Website and Online Ordering System: What It Has to Do in 2026

Your own website is the only ordering channel where you keep the whole ticket. Every order routed through a marketplace gives away 15-30% of the subtotal — and gives away the customer with it. You never learn their name, their phone number, or what they ordered last time.

Marketplaces aren't the enemy — they're paid acquisition, and you should stay on them. But the two channels have such different economics that the decision deserves a calculator, not a vibe.

The Four Jobs a Restaurant Website Has to Do

Everything else is decoration.

1. Get found — in organic results, in the map pack, and increasingly inside AI-generated answers.
2. Be machine-readable — structured enough to qualify for rich results and to be quoted accurately by an answer engine.
3. Load fast on a phone — functionally the only device that matters for this traffic.
4. Take an order and take money — on your domain, through your processor.

A brochure site with a hero photo, a PDF menu, and a Facebook link fails all four.

Kill the PDF menu

Fix this before anything else. It fails on every axis at once: a PDF isn't page content Google can treat as menu text, an answer engine asked "do they have gluten-free options" can't read it reliably, and on a phone it's a pinch-and-zoom exercise that makes people leave. An inaccessible menu — especially a scanned one, which is just a picture of text — is also the most commonly cited defect in restaurant web accessibility complaints.

Your menu must be HTML text on the page. Item names, descriptions, and prices as real text a browser can reflow, a screen reader can read, and a crawler can index.

Be careful what you read elsewhere on the legal side. There is no federal regulation setting a technical accessibility standard for private restaurant websites — the 2024 DOJ rule adopting WCAG 2.1 AA applies to state and local government entities under Title II, and plenty of vendor blogs conflate the two. What is true: courts have found restaurant websites covered by ADA Title III where they connect to a physical location (Robles v. Domino's, 9th Circuit), plaintiffs treat WCAG 2.1 AA as the working benchmark, and some states add their own exposure. Practically: real HTML menu, captions on anything with audio, keyboard-navigable checkout. Cheap to build in, expensive to retrofit. Source: ADA.gov

The Commission Math

What the marketplaces actually charge

Every major platform is tiered, and the tier changes your visibility in the app, not just your rate. "DoorDash takes 30%" is wrong often enough to be worth unlearning.

Uber Eats raised marketplace commissions in March 2026 — its first increase in roughly a decade — moving Lite from 15% to 20%. Uber One orders on the Plus plan bill at 30% rather than 25%; self-delivery sits at 15%. Grubhub's public plan pages are less specific than DoorDash's, so treat those figures as approximate. Source: DoorDash · Source: Restaurant Dive · Source: Restolabs

Headline commission isn't the whole cost. Sponsored listings you bid on, promotions you fund to stay visible, and refunded orders you eat all push the effective rate above the tier rate on your contract — ghost kitchen and delivery economics works through that stack and how to measure your own. Most operators respond by listing marketplace prices 10-20% above in-store — so the guest already pays more to order the convenient way, they just don't see it itemized. Source: Sauce Note that a 10-20% bump doesn't actually offset a 25% take: to net your in-store price after a 25% cut you need a 33% markup, because you're solving for the price that survives the cut, not one that offsets it.

Jurisdiction check: Commission caps are city-level and unstable. New York City and San Francisco made theirs permanent; Los Angeles, Seattle, Chicago, Washington DC, and Las Vegas ran temporary ones. NYC caps delivery at 15%, other services at 5%, and card processing at 3% — but a 2025 settlement and follow-on legislation let platforms add an optional "enhanced services" tier on top. If your city caps commissions, check the ordinance and your platform's local rate card, because the two don't always agree. Don't treat a cap as protection.

What direct ordering actually costs

Three line items, and only one scales the way a commission does.

• Subscription: $0 (ordering bundled into a POS you already pay for) up to roughly $500/month for a full platform.
• Payment processing: about 2.9% + $0.30 per online transaction — the published Stripe and Square rate, a benchmark rather than a quote. Source: Stripe
• Courier dispatch, only if you deliver: Uber Direct or DoorDash Drive at roughly $6-$9 per delivery (Uber Direct around $6.99 through some POS integrations). Source: Restaurant Business

Pickup and catering orders on your own site carry almost no marginal cost beyond processing. That's why pickup-heavy concepts — pizza, wings, sandwiches, anything with a counter — get the biggest win.

The worked example

400 online orders a month at a $38 average ticket — $15,200 in monthly online sales — taken three ways.

The number that changes minds isn't the total, it's the breakeven. At a $38 ticket a 25% commission is $9.50 per order, so roughly 53 orders a month — under two a day — moved from the marketplace to your own site pays for a $500/month platform outright. Everything past that is margin you keep.

Be strict with that number, though: it's gross of processing. Each $38 order you move still costs about $1.40 to run on your own site, so the honest breakeven is nearer 62 orders — still two a day. Run it on your own ticket before believing any version of it; at an $18 average the same commission nets you only $3.68 per moved order, so the same platform needs roughly 136 orders a month — closer to five a day — to clear. If you're unsure what a direct order is worth after labor and packaging, work through restaurant financial basics first.

The honest counter-argument

Don't delist. Marketplaces produce genuinely incremental orders from people who've never heard of you, and paid acquisition with a knowable cost is the right mental model. A 25% commission on a stranger's first order is a fair acquisition cost; on a regular's eleventh order it's a leak.

So: dual-channel, with deliberate conversion.

• Bag inserts in every marketplace order, carrying a real offer rather than a logo.
• QR codes on receipts and packaging, pointing at your ordering page.
• A first-order discount on the direct channel, funded by the commission you're not paying.
• Don't mark up your own channel. The strongest lever, and the one most operators get backwards. If your prices are the honest in-store prices and the marketplace's are 15% higher, direct is visibly cheaper and you never have to argue the point.

One caveat on that last item: several direct-ordering platforms charge the guest a service fee. If yours adds 5% at checkout, "order direct, it's cheaper" is weaker than you think. Ask what the customer sees on the final screen, then go look at a live one.

Getting Found

Your Google Business Profile is doing more work than your website

Local marketing for a new restaurant covers claiming and completing the profile step by step, along with reviews and launch week; this section covers only how it relates to your site.

With a physical address, your Business Profile plus review volume drives most local discovery; the website's job is to be the destination it points at, and the preferred ordering link on it. Google retired in-Google food ordering transactions — "Order online" now surfaces a list of ordering links, direct and third-party together, and a business can designate a preferred ordering partner carrying a "Preferred by business" badge. That setting takes five minutes, has direct revenue consequences, and most operators have never opened it. Source: Google

What actually changes for AI search

Less than the panic suggests. Google's stated position is that best practices for AI Overviews and AI Mode are the same as for Search — indexable content, Googlebot not blocked, structured data, good page experience, unique content, a current Business Profile — with no additional requirements. And ignore anyone selling an AI package built on llms.txt: Google has said Search doesn't use it and isn't planning to. Source: Google Search Central · Source: Search Engine Journal

The leverage is in what that guidance doesn't cover: assistants outside Google lean on licensed review and place data — ChatGPT on Yelp and Foursquare, Perplexity on Tripadvisor. So the highest-value "AI optimization" for a restaurant is unglamorous — consistent name, address, and phone across the major directories, real review volume, a current rating.

The on-page half is no more glamorous. Write the answers people actually search for as plain text: hours by day, parking, whether you deliver to a named neighborhood, whether there's a patio, the gluten-free options, catering. Plain declarative sentences get quoted; marketing copy doesn't.

Structured Data: The Markup That Matters

Restaurant is a subtype of LocalBusiness and it's the type to use, in JSON-LD. Required: name and address as a nested PostalAddress. Worth the twenty minutes: telephone, url, openingHoursSpecification, servesCuisine, priceRange, geo, image, and menu as a fully-qualified URL; multiple locations use department. Two rules get sites in trouble — name, address, and phone in the markup must match your Google Business Profile exactly, and everything in the markup must also be visible to a human on the page. Source: Google

Two caveats the vendor blogs skip. Menu and MenuItem markup helps machines understand your menu but is not a standalone rich result type, so don't expect a visible payoff. And aggregateRating and review markup are meant for sites capturing third-party reviews — a restaurant marking up its own hand-picked testimonials is asking for a manual action.

One more, since half the internet says otherwise: structured data does not improve rankings. It affects eligibility for rich results and machine comprehension. Validate in the Rich Results Test, then watch the Search Console enhancement reports — that's where you find out it broke after a template change.

Speed on a Phone

LCP under 2.5s, INP under 200ms, CLS under 0.1, judged at the 75th percentile of real Chrome users over a rolling 28-day window. That last clause matters: it's field data from actual visitors, not a lab score you can pass by testing on office wifi. Source: web.dev

Restaurant sites fail in four predictable places: an autoplaying hero video, which wrecks LCP on cellular; uncompressed food photography straight off the camera at 6MB a frame; a stack of third-party widgets each loading its own JavaScript (reservations, chat bubble, review badge, three pixels); and an ordering system in an iframe that blocks interaction while it boots, which is an INP problem specifically. Test the ordering page, not the homepage — the homepage is the one you look at, the ordering page is the one that carries the money, and it's usually the slowest on the site.

Taking the Order

What first-party ordering needs in order to stop losing orders you already earned: an HTML menu with modifiers and out-of-stock toggles staff can flip themselves; pickup and delivery; scheduled and order-ahead; tipping; Apple Pay and Google Pay, so a first-time guest never types a card number on a phone; SMS confirmation; and catering or large-order handling if that's a real part of your business.

Then the point most guides skip. The real cost of direct ordering is order routing. If direct orders land on a separate tablet nobody watches during a Friday rush, they get missed — and a missed direct order costs more than a 30% commission ever did, because you lose the ticket, the customer, and the review. Evaluate POS and KDS integration harder than you evaluate design.

The same is true physically. A direct channel creates a staging problem the marketplaces used to own for you: bagged orders going cold while the guest is still parking. The fix is a real pickup station, not a corner of the expo line — heated holding cabinets and pickup shelving so hot food holds temp between bag-up and pickup, and a dedicated expo and packing table so packing doesn't compete with plating for counter space. Skipping this is why some operators decide direct ordering "doesn't work": the economics were fine, the execution wasn't. Before you push volume at the line, check your throughput assumptions against the complete equipment checklist for new owners.

Then the compounding asset. A direct order gives you a name, phone, email, and order history. A marketplace order gives you none of that. That list is the entire reason email and SMS marketing works for restaurants, and it's why the direct channel gets cheaper per order every year while the marketplace channel never does.

Four things to settle before launch, not after. Consent: texting your list requires express written consent captured at checkout with a working opt-out — federal rules, with state additions. Get the checkbox right at build time; a list you can't legally text is a total loss. Alcohol: to-go cocktails and beer or wine delivery vary by state and sometimes by county, and pandemic-era allowances expired unevenly — confirm with your state ABC or liquor authority first. Sales tax: marketplace facilitator laws mean the platform often remits tax on marketplace orders while you remit it on direct ones. Ask your bookkeeper or state revenue department how direct orders are treated before your first month closes. Menu labeling: federal calorie rules apply to chains at 20+ locations, so almost certainly not you — but some jurisdictions set their own allergen-notice requirements, and one call to your health department settles it.

Is a Branded App Worth It?

An app does not acquire customers — nobody browses the App Store looking for dinner. An app retains customers, and the feature you're actually buying is the push notification: free to send, lands on the lock screen, never sees a promotions folder. So it's worth it above a repeat-order threshold. If a meaningful share of your guests order more than twice a month, push is a real revenue lever; if you have forty regulars and no loyalty program, it's a line item that does nothing. One thing shifts the math — most bundled platforms include a branded app in the subscription, which is very different from commissioning one.

Choosing a Platform

Ignore the feature grids. These nine questions decide whether you're happy in year two.

Three categories are worth running through it: POS-native ordering (Toast, Square) if your POS is already chosen and integration matters most; commission-free ordering specialists, the bundled-platform category; and a DIY build on a CMS with an ordering plugin, cheapest in cash and most expensive in operator time.

We earn a commission if you sign up through this link, which is exactly why you're getting the tradeoff along with the pitch. In the bundled category, Owner.com is where we'd point a first-time operator: month-to-month with no long-term contract and no cancellation fee, a flat-rate tier that removes per-order restaurant commission entirely, and website, ordering, branded app, loyalty, and email/SMS bundled rather than stitched together — which matters more than it sounds, because every integration seam is a place orders get lost. The tradeoffs, in the same breath: the Flex plan is $249/month plus a 5% restaurant fee per order, Flat-Rate is $499/month with no per-order restaurant fee, and both plans charge the guest a 5% order support fee. Your customer is not paying zero fees on your own site, which directly undercuts the "direct is cheaper" pitch this whole guide is built on. Know that before you sign, not after a regular asks you about it. Source: Owner.com pricing

The plan arithmetic is head-math: the $250/month gap between the two is covered by 5% of about $5,000 in monthly online sales. Below that Flex is cheaper; above it Flat-Rate, and the gap widens fast. Don't buy the bigger plan on projected volume.

If you pick something else, run it through the table anyway. The evaluation is the deliverable here; the recommendation is one answer to it.

The Ghost Kitchen Version

If you're delivery-only, read this section first. Your website isn't marketing — it's the entire storefront. No sign, no window, no walk-by: every customer arrives through a screen. The margin structure underneath it — effective take rates, delivery pricing, and what a delivery-only concept has to charge to clear — is ghost kitchen setup and delivery economics; this section is the site itself.

Your Google Business Profile works differently. Google has specific guidance for virtual and delivery-only brands: a brand in a shared kitchen with no customer pickup must hide its address and register as a service-area business with defined delivery areas, while co-located brands that do offer pickup need distinct branding and permanent separate signage to hold a storefront listing. Verify against Google's current guidelines before you file — this policy area moves. Source: Google

Treat that as a planning input, not a complaint. A hidden address means weaker map-pack presence, so your site and the marketplaces carry more of the discovery load — and a concept that can offer counter pickup gets a materially stronger listing than one that can't. Worth knowing while you're still choosing a space.

Multiple virtual brands means multiple sites, each with genuinely distinct content. Three template clones with swapped logos is duplicate thin content and will rank like it. If you can't write three real menus and three real about pages, don't launch three brands — and the platforms enforce the same thing on their side, with published differentiation and performance minimums the ghost kitchen guide lists in full.

Converting marketplace customers to direct is the whole game. A storefront gets some of that done free by foot traffic and signage. You get neither, which makes the bag insert your entire retention program.

Then check capacity. A direct channel that works produces volume the line has to absorb, and delivery menus concentrate on a few stations. The fry station is the usual throughput ceiling — fryers and grills are where a delivery-heavy menu hits a wall first — and scheduled and catering orders need cold staging, which means refrigeration sized for orders that exist hours before they're cooked. For the equipment side of a delivery-only build, the ghost kitchen equipment guide covers what to buy and skip.

What It Costs to Stand Up

Ranges, not quotes.

Two line items dominate all three and neither is code: photography and menu copy. They're the first things first-time operators cut and the first they regret. Budget them inside your startup cost estimate, and if money is tight make the trade on the equipment side instead — where to buy used commercial kitchen equipment can move five figures out of the buildout, and if the problem is timing rather than budget, run the equipment financing math.

Frequently Asked Questions

Do I need my own website if my restaurant is already on DoorDash and Uber Eats?

Yes, for arithmetic reasons rather than principle. A marketplace order costs 15-30% of the subtotal and gives you no customer information; a direct order costs a subscription plus roughly 2.9% + $0.30 in processing and gives you a name, phone, email, and order history. At a $38 ticket on a 25% tier, moving about 62 orders a month to your own site pays for a $500/month platform. Stay listed for discovery — just stop paying acquisition rates on customers you already acquired.

How much commission do delivery apps charge restaurants?

It depends on the plan tier, and the tier changes your in-app visibility as well as your rate. DoorDash runs 15%, 25%, and 30% on marketplace delivery with 6% on pickup. Uber Eats runs 20%, 25%, and 30% after its March 2026 increase, pickup at 7%, and Uber One orders on the Plus plan at 30%. Grubhub's marketing commission is roughly 5% to 20% by plan with delivery from around 10% on top. Effective rates land higher once processing, promotions, and refunded orders are counted.

Is commission-free online ordering actually free?

No, and it's the question to put to a salesperson directly. "Commission-free" usually means no per-order fee charged to the restaurant — real and valuable, but you still pay a subscription, roughly 2.9% + $0.30 in processing, and $6-$9 per delivery if you dispatch couriers. Several platforms also charge the guest a service fee at checkout, often around 5%, so your customer isn't paying zero either. Open a live checkout on that platform and look at the final number a guest sees.

What schema markup does a restaurant website need?

Use the Restaurant type, a subtype of LocalBusiness, in JSON-LD. Required are name and address as a nested PostalAddress; worth adding are telephone, url, openingHoursSpecification, servesCuisine, priceRange, geo, image, and menu as a fully-qualified URL. Name, address, and phone must match your Google Business Profile exactly, and everything in the markup must be visible on the page. Skip aggregateRating and review for your own testimonials — that markup is meant for third-party review capture, and misusing it risks a manual action.

How do I get my restaurant to show up in ChatGPT and AI search results?

For AI Overviews and AI Mode, Google says there are no additional requirements beyond normal SEO: indexable content, Googlebot not blocked, structured data, good page experience, a current Business Profile. Assistants outside Google lean on licensed review and place data — ChatGPT on Yelp and Foursquare, Perplexity on Tripadvisor — so the highest-leverage work is consistent name, address, and phone across the major directories, real review volume, and a current rating. On your own site, write plain-text answers to the questions people actually ask; plain sentences are what gets quoted.

Does a ghost kitchen need its own website?

More than a storefront restaurant does. With no signage and no walk-in traffic the site is your only storefront, and your Google Business Profile is weaker by design — a delivery-only brand in a shared kitchen has to hide its address and operate as a service-area business, which cuts map-pack presence. That pushes discovery onto the marketplaces and your own site, and makes converting marketplace customers to direct the central task. Bag inserts and receipt QR codes are your only physical touchpoint, so treat them as a retention program, not packaging.