Find creators yourself from tagged posts and neighbourhood tags, score them in a sheet with an AI assistant, and give each a one-page brief that requires a clear "Ad" label. Track each comped meal with a unique code word and booking link, then compare its cost with the gross profit on tracked bookings after 14 days.
The AI shortlists and drafts; it cannot find accounts, verify followers or decide what is lawful. The rule that matters most is that a free meal given for a post is an incentive, so the post must say it is an ad, and the restaurant shares responsibility for that. Everything else is arithmetic and housekeeping.
What a comped meal really costs, and how many bookings repay it
Start with the sums, because they decide everything after. A comped meal costs less than its menu price and more than nothing. Suppose a meal for two is worth $120 on the menu and your food and drink cost is 30%, so the ingredients and drinks cost $36. Add 15 minutes of host or manager time at $25 an hour, and a share of your own planning time. If you spend two hours on the whole exercise and host three creators, that is $16.67 each. Every figure here is illustrative; swap in your own.
| Cost line | How it is worked out | Amount |
|---|---|---|
| Ingredients and drinks | 30% of the $120 menu value | $36.00 |
| Host time | 15 minutes at $25 an hour | $6.25 |
| Planning share | 2 hours at $25, split across 3 creators | $16.67 |
| Cost per creator visit | $58.92 |
What a tracked booking returns is gross profit, not takings. If the average booking is 2.8 guests spending $31 each, that is $86.80, and at the same 30% cost it leaves $60.76. So one tracked booking roughly repays one comped meal ($58.92 divided by $60.76 is 0.97). Staff are paid for an ordinary service whether or not the table is comped, which is why this short-term view leaves them out.
Two things move the break-even point. The first is a fee: a creator who also charges $150 needs ($58.92 plus $150) divided by $60.76, which is 3.4, so four tracked bookings. The second is displacement. If the comped table would otherwise have sold at 20:00 on a Friday, you also lose $84 of gross profit ($120 at 70%), the cost rises to about $143 and the break-even to three bookings. That is why comped visits belong in your quietest service and not your busiest.
Finding creators an AI cannot find for you
A chat assistant has no reliable live view of platform accounts. Asked to "suggest local food influencers", it produces plausible handles, some real and some invented, with follower counts nobody checked. Use it to rank accounts you have found, never to find them. Four places to look, in order of warmth:
- Guests who already tag you. Posts that mention or tag your restaurant are the warmest list, because those people chose you unpaid.
- Neighbourhood tags and neighbouring venues. Who posts from the café two doors down, the market or the park? Who gets tagged by other restaurants nearby?
- Your staff and regulars. Ask "who do you follow for places to eat round here?" at the pass or in your next guest email.
- Creators who message you. Treat the cold pitches the same way: into the sheet, scored on the same columns.
For each candidate, record the handle and link, followers, average likes and comments on the last nine posts, how often they post about food, whether past partnerships were labelled and what their audience looks like. For audience, ask the creator for a screenshot of the location breakdown in their own account insights; the label differs by platform, so ask them what they see. An audience that lives elsewhere cannot fill your Tuesday tables, however large it is. A dozen candidates take about 45 minutes to collect.
Scoring the shortlist: a sheet, a prompt and an arithmetic check
Fill the sheet by hand from public profiles. Here are five candidates, with engagement worked out as average likes plus average comments, divided by followers (illustrative numbers):
| Candidate | Followers | Avg likes | Avg comments | Engagement | Local audience | Notes |
|---|---|---|---|---|---|---|
| A | 9,400 | 610 | 38 | 6.9% | 64% | Posts neighbourhood food most weeks |
| B | 22,000 | 410 | 9 | 1.9% | 12% | Travel and lifestyle; audience elsewhere |
| C | 3,500 | 240 | 31 | 7.7% | 71% | Student account; posts weekly |
| D | 41,000 | 95 | 4 | 0.2% | Not shared | Large following, almost no response |
| E | 6,200 | 330 | 52 | 6.2% | 58% | Lunch spots, mostly short video |
I promise no benchmark here, because a published average for "good engagement" would be a guess. Compare the candidates with each other and with what your own first campaign returns. Then hand the table to an assistant:
You are helping a 40-seat restaurant choose creators for comped visits. Below is a
sheet I filled in by hand from public profiles. Engagement = (average likes +
average comments) / followers. Calculate it yourself and show the working.
Rank the candidates for a restaurant that needs local covers Tuesday to Thursday.
Do not judge whether followers are real; you cannot see that. For the top three,
tell me what to ask each before inviting them.
[paste the table]
A typical reply (illustrative) ranks C first, then E, then A, and then B and D, with questions such as "How many of your recent restaurant posts were unpaid?" and "Can you share your audience location screenshot?". It also contains three things to correct:
- An arithmetic slip that changed the order. It gave E an engagement of 7.1%. The correct sum is 330 plus 52, which is 382, divided by 6,200: 6.2%. That puts A (6.9%) above E, so the ranking becomes C, A, E. Always recompute the figures that decide the order.
- "A's audience looks authentic". A model cannot see follower quality. Delete the sentence, and do your own check of the comments for repeated generic replies.
- "D has the largest reach". Followers are not reach, and D's response is the weakest on the sheet. Ignore it.
The shortlist is C, A and E. B is skipped because its audience lives elsewhere, and D because it did not share insights and its response is close to nil.
Disclosure: what ad regulators and the platforms ask for
Name the rule before writing the brief, because it shapes the brief. Advertising regulators publish influencer guidance for creators and for brands, and the guidance in the larger markets reads alike where it matters. It is guidance rather than legal advice, so check your advertising regulator's disclosure rules for each market your audience is in. The common ground starts with the incentive. Whenever a brand gives a creator a payment or any other incentive, a free meal included and whether or not the creator asked for it, content featuring the brand has to be obviously identifiable as advertising. A clear "Ad" label at the start is the safest form. A label buried in a list of hashtags at the end, or visible only after a "See more" tap, is not enough, and vague wording such as "gifted", "thanks to X for making this possible", "sp", "spon" or "collab" is unlikely to count as clear.
The rest of the common ground covers placement and responsibility. The disclosure should sit with the endorsement and be hard to miss, and in a video it belongs in the video itself, not only in the description. A hashtag such as #ad is fine but not required. A platform's own paid partnership label may not be enough on its own, so use it alongside your own "Ad" label rather than instead of it. Brands share the duty with creators: published guidance expects advertisers to train and monitor the people promoting them, and points out that it is far easier to review posts before they go live than to hunt for them afterwards. Two more points matter for a comped meal. You cannot make an incentive conditional on a review being positive, and an endorsement must reflect the creator's honest opinion.
Put that together and the brief has five requirements: "Ad" as the first word of the caption, the platform's label as well, an on-screen "Ad" in the first seconds of any video, no hiding the label among hashtags, and no demand that the opinion be positive. On Instagram, Meta's branded content policies treat content made for an exchange of value, such as payment or free gifts, as branded content, and require the creator to tag the business partner with the branded content tool, which adds the paid partnership label. The first time a creator tags your restaurant, Instagram sends you a request, and tagged partners get the post's insights. The Instagram paid partnership label page has the steps, and your account needs to be a professional one. On TikTok, creators turn on Disclose commercial content under Content disclosure and ads and choose Branded content, which shows a Paid partnership label; TikTok's disclosure setting guide lists the steps. For how another sector handles this in practice, see the skincare influencer disclosure tutorial.
One extra rule is easy to forget. Never pair a comped meal with a request for a Google or other review. Google's review policy bars incentives for posting reviews, and a free meal is one.
A one-page creator brief you can copy and adapt
A brief has one job: remove the guessing that makes posts inaccurate or unlabelled. Ask an AI assistant to turn your decisions into a page, then edit it. This is what a finished one looks like (illustrative):
BRIEF: [restaurant name] and [creator handle]
VISIT: Tuesday 17 November, 18:30, table for two, window table held. Ask for [host].
WE PROVIDE: two starters, two mains, one dessert to share, two soft drinks. Anything
extra you order is yours to pay. Staff service charge is covered by us.
WE ASK: one Reel of 15 to 40 seconds and three Stories, posted within 7 days of the
visit. Tag @[restaurant handle] and add our location.
DISCLOSURE (required): start the caption with "Ad: meal provided by [restaurant]".
Use Instagram's paid partnership label too. Show "Ad" on screen in the first
seconds of the Reel and on each Story frame that mentions us. Keep the label out of
the hashtags. Use at most five hashtags.
YOUR OPINION: say what you honestly think. We only ask that facts are right (dish
names, prices, opening hours) and that you send the draft 48 hours before posting
so we can correct facts. We will not ask you to change an opinion.
PLEASE AVOID: allergen or dietary claims unless the host has confirmed them; "best in
town" claims we cannot back up; filming other guests' faces.
NO REVIEWS: please do not post a Google or other review as part of this arrangement.
RIGHTS: we may reshare your post on our own accounts with credit. Any other use
(website, ads, print) only with your written OK.
TRACKING: your booking link is [link with tags]. Guests can also say the word
HARVEST when they book. We will share how many bookings used either, 14 days after
you post, and we would like a screenshot of the post's insights at day 7.
Three choices in it matter. First, the draft review is for facts and the label, not for taste; the more you script a post, the more it is your advertising, whoever's name is on it. Second, the "no review" line protects you under Google's rules. Third, the rights line stops a pleasant visit turning into an argument about reusing the video. If you might want to boost the post as a partnership ad, say so, since the creator has to allow it.
Three outreach drafts, and what to fix in them
Ask the AI for short messages in your own voice, and feed it the brief so the promises match. Three cases cover most of what you will send. A cold message to a local creator might come back like this (illustrative):
Hi [first name], I run [restaurant] and have followed your food posts for a while. We would like to invite you for a meal for two on Tuesday 17 November, with nothing to pay. All we ask is one Reel and a few Stories, labelled as an ad, and you say what you honestly think. The one-page brief is attached; tell me if the date does not suit and I will offer another. Thanks, [your name]
It is short, it names the date and the obligation, and it mentions the label in the first message, which saves an awkward conversation later. Check that it does not say "we love your content" without naming a post. Generic flattery reads as a mass message. When a creator replies asking for a fee, you can decline politely without losing the door:
Thanks, [first name]. For a first visit we only do the meal, because we measure each visit against bookings. If the post brings tables, I would be glad to talk about a paid piece afterwards.
And for the creators you do not choose, a decline needs only two sentences: thanks for the offer, not this time, and nothing that sounds like a promise. A model will want to add "we will definitely keep you in mind". Remove it.
Hosting the visit so the post is accurate
Most errors in influencer posts start at the table. Brief the team before the visit with a short list: who the guests are, what is comped, who covers service charge, which dishes they will be served, and who talks to them about allergies. Ask about allergies and dietary needs before ordering, in exactly the way you ask any guest, and let the kitchen's record, not the host's memory, decide what can be claimed.
Seat them somewhere with good daylight or light you control, and agree up front what they may film. Kitchen footage needs the cook's consent, and other guests' faces are best left out of frame. Spend 15 minutes with them, not the whole meal, and do not stand over the table. The comp is the cost; your behaviour is what the post will reflect. Finally, check your Google Business Profile the day before the post goes live. A guest who sees a creator's Reel and then finds the wrong hours on Google will not book.
Tracking bookings: code word, link and a 14-day window
You need two ways of catching a booking, because no single method catches them all. The first is a code word. Choose a plain word spelled one way, such as HARVEST, rather than the creator's name, which guests spell three ways. Put it in the brief and ask hosts to log it. The second is a tagged booking link. Google Analytics' help says to always use the source, medium and campaign parameters and to keep values lowercase, so a link ends with ?utm_source=instagram&utm_medium=creator&utm_campaign=nov-creators&utm_content=creator-a. The Google Analytics page on campaign parameters explains them. Some booking widgets drop the tags, so click your own link on a phone, make a test booking and see whether the source survives.
If your booking form has a free-text field, add "How did you hear about us?". Then fix the window before the post goes live. Fourteen days after posting is my default, not a rule, and what matters is that you choose it first. Keep a sheet with one row per creator. This one is filled in (illustrative):
| Creator | Visit | Posted | Code | Tracked bookings | Covers | Window closes |
|---|---|---|---|---|---|---|
| A | Tue 17 Nov | Fri 20 Nov | HARVEST | 5 | 14 | Fri 4 Dec |
| C | Wed 18 Nov | Sat 21 Nov | LANTERN | 3 | 8 | Sat 5 Dec |
| E | Thu 19 Nov | Mon 23 Nov | COPPER | 1 | 3 | Mon 7 Dec |
The insights screenshot at day 7 shows reach, saves and link taps. They tell you whether the post was seen, not whether it earned money, so keep them out of the return sum. Anything that does not use a code or link, such as a guest who simply walks in, goes in a separate "unlinked" column. If you count it as a return you will flatter every campaign. For a general look at which numbers deserve a place on a monthly sheet, see AI KPIs for small businesses.
A three-creator test, with illustrative numbers
The 40-seat restaurant in this example, open Tuesday to Sunday, invites C, A and E for comped meals at quiet services. The cost and margin assumptions are the earlier ones: $58.92 per visit and $60.76 of gross profit per average booking. Fourteen days after each post, the tracked results (illustrative) are in:
| Creator | Tracked bookings | Covers | Spend at $31 a head | Gross profit at 70% | After the $58.92 cost |
|---|---|---|---|---|---|
| A | 5 | 14 | $434 | $304 | +$245 |
| C | 3 | 8 | $248 | $174 | +$115 |
| E | 1 | 3 | $93 | $65 | +$6 |
| Total | 9 | 25 | $775 | $543 | +$366 |
Total cost was $177: three visits at $42.25 each for ingredients and host time, plus $50 of planning. Gross profit on tracked bookings was $543, so the test returned about three times its cost. A and C clearly paid. E only broke even, so one visit does not justify a second until something changes.
The decisions follow from the numbers, not from the follower counts. Re-invite A and C for a different service, thank E, and hold any talk of fees until a creator has shown at least three tracked bookings from a free visit. Keep the caveats in view: nine bookings is a small sample, a different Tuesday could have gone differently, and the sheet does not count guests who came without using a code. Treat it as a reason to run the next round, not a forecast. Once a creator's guests arrive, the no-app loyalty scheme is a cheap way to turn a first visit into a second.
An illustrative slip shows why the displacement line matters. Imagine the first creator is comped at 20:00 on a Friday, at a table that had been requested twice earlier in the day. Two paying guests are turned away, so the visit costs the $59 plus $84 of lost gross profit, and a single tracked booking afterwards does not cover it. The cost only shows up when you ask who would have sat there. Moving the next visit to a Tuesday fixes it.
Checking a creator's draft before it goes live
When the draft arrives 48 hours ahead, check it against a list, not by feel. An AI assistant is good at this kind of comparison if you give it the list and the facts and tell it not to rewrite opinions:
You are checking an influencer's draft post for a restaurant that gave the creator a
free meal. For each item, return pass or fail, quote the exact words, and suggest a fix.
Do not rewrite the creator's opinions.
1. Does the caption start with "Ad" before any "more" cut-off?
2. Is the label separate from the hashtags?
3. Is "Ad" on screen in the first 3 seconds of the video? (I describe the video below.)
4. Any allergen or dietary claim?
5. Any claim such as "best", "cheapest", "healthiest" or "award-winning"?
6. Do dish names, prices and hours match the fact sheet?
7. Is a Google or other review requested or implied?
FACT SHEET: [dishes, prices, hours]
VIDEO: [describe the first 3 seconds and any on-screen text]
CAPTION: [paste]
A typical result (illustrative) fails three items: the caption opens with "Date night at [restaurant]" and carries "#ad" last, after four other hashtags; it calls the lasagne "the best in town"; and it gives a price of $16 where the fact sheet says $18. Send the creator that list, not a rewritten caption, and let them fix their own words. If they decline to label the post as an ad, do not accept it; the label is a condition of the meal, and under advertising regulators' guidance the brand carries responsibility as well.
When the post is live, check it once more against the same list, note the post date in your tracking sheet and start the 14-day clock. The wider picture of where this sits among your other channels is in the four-week AI marketing plan for restaurants.
Questions restaurant owners ask before inviting creators
Do I have to tell creators to label a free meal as an ad?
Yes. Advertising regulators' influencer guidance treats a free product or meal as an incentive that must be disclosed, and expects the label to be obvious. Put it in the brief, ask for "Ad" at the start of the caption plus the platform's paid partnership label, and check the draft before it goes live. This is guidance, not legal advice, so check your advertising regulator's disclosure rules for the market your audience is in.
Should I pay creators a fee or only comp a meal?
Start with a comped meal in a quiet service and judge by tracked bookings. On the example numbers a $150 fee needs about four tracked tables to repay, against one for a comped meal alone. Pay a fee only for a creator whose free visit already brought several bookings, and agree the label and the facts in writing first.
Can I ask the creator to leave a Google review as well?
No. Google's review policy bars incentives for posting reviews, and a free meal is an incentive. Keep the arrangement to the creator's own post, labelled as an ad, and ask for nothing on Google or other review sites. Guests who book through the post can review you later on their own, like anyone else.
Can AI find local influencers for me?
It can rank accounts you have found, but it has no reliable view of platform accounts and may invent plausible handles and follower counts. Find candidates from posts that tag you, neighbourhood tags and what your staff and regulars follow, copy their public numbers into a sheet, and then ask the AI to score and compare them.
Further reads
- How to Make TikToks and Reels for a Small Business With AI — Short video you can make yourself while creators post theirs.
- 7 AI Mistakes Restaurant Owners Make With Bookings and Reviews — Booking and review slips to avoid while a post sends new guests.
- Restaurant Gift Card Programme: Terms, Offers and Tracking With AI — Another offer to hand to a guest who arrives through a post.
- Restaurant No-Shows: Deposit Rules by Party Size, AI Confirmations — Protect tables when a post sends a burst of bookings.
- Write a Good Bio for a Restaurant Owner With ChatGPT: About Page and Press — An interview-style ChatGPT prompt turns a restaurant owner's own facts into three bios, for the About page, press requests and social, with no invented awards.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: advertising regulators' published guidance for influencers and brands on labelling ads and disclosing free products; Meta's branded content policies and Instagram's help page on the paid partnership label; TikTok's help page on the commercial content disclosure setting; Google's review policy on incentives; Google Analytics help on campaign URL parameters (checked October 2026).