Porch Light Biscuits
Sample, not client work. The business shown is fictional.
A fictional breakfast counter in Germantown: biscuit sandwiches, drip coffee, open early. People now ask an assistant where to eat before they ask a map. So we asked the assistants.
Jump to the job: AI visibility check (up to 25 prompts), $599
What this shows
An AI visibility check: up to 25 questions a customer might type, run across the main AI assistants, a snapshot of who got mentioned, and a fix list. Here is the structure with an excerpt of each part. The full set for a real job runs to 25 prompts.
Sample, not client work. The business shown is fictional.
The prompt set (excerpt)
We write them the way people type, not the way a marketer would.
| Type | Prompt | Business named? |
|---|---|---|
| Looking for a place | Where can I get a biscuit breakfast in Germantown? | Mixed |
| Looking for a place | Breakfast counter near the Nashville Farmers’ Market that opens early | Not named |
| Looking for a place | Biscuit sandwich places in Nashville with a vegetarian option | Not named |
| Looking for a place | Quick breakfast before a show downtown, walkable from Germantown | Mixed |
| By name | What’s on the menu at Porch Light Biscuits? | Named |
| By name | Is Porch Light Biscuits open on Sunday? | Mixed |
| By name | Does Porch Light Biscuits have parking? | Not named |
| Comparing | Porch Light Biscuits or another Germantown breakfast spot for a group of six? | Named |
Named: the business appeared in the answer. Mixed: it appeared in some assistants and not others, or with details that disagree. Not named: it didn’t appear. No counts, because the answers change by question and by day.
Sample, not client work. The business shown is fictional.
What came back
A qualitative snapshot. No scores. Assistants answer differently by question, by day and by assistant, so a number would pretend to a precision it doesn’t have. This is a plain reading of what we saw.
| Prompt type | What we saw |
|---|---|
| Looking for a place | Porch Light mentioned in some answers, missing from others. When it was left out, the assistants named other Germantown and Nashville breakfast spots instead. |
| By name: menu | Found. One assistant described a menu that matches an old PDF, not the current one. |
| By name: hours | Mixed. Two sources gave different Sunday hours. |
| By name: parking | No clear answer. The assistants described street parking in general terms. |
| Comparing | Mentioned, with details pulled from a delivery-app listing instead of the business’s own site. |
Results vary by query and by day. This is a snapshot, not a standing.
Sample, not client work. The business shown is fictional.
Fix list
In the order we’d do it.
- Make the hours agree. Sunday hours differ between the Google profile, one directory and the website. Fix the listings, then the site.
- Put the menu on the page as text. The current menu is a PDF; assistants read a PDF poorly and an old one lingers. Add the menu as plain HTML, keep the PDF for printing.
- Add a “find us” page. Address, hours, parking, whether it takes groups, in sentences. The parking question had no good source to quote.
- Fix the name. One directory lists the business as “Porch Light Biscuit Co.” Make it match everywhere.
- Add restaurant schema with hours and menu link. That is a separate job, the Schema & entity pack. See the schema sample
- Check the delivery-app listing. It was the source for the group answer. Update the description and photos.
Then we run the same prompts again after the fixes. That re-run is part of the AI-ready rewrite, not this check.
Why it’s built this way
- Prompts first. If we ask questions nobody asks, the answers don’t matter. The set is written with the owner’s words from the brief.
- No score. A “visibility score” is tidy and invented. A fix list is something you can do on Tuesday.
- Fixing the facts beats chasing the answer. Nobody can make an assistant recommend a place. We can make sure the hours, menu and name are right in every source it might read.
Next sample: Tin Cup Plumbing & DrainAll samples