Method
Published information about recurring restaurant specials is mostly stale. Listicles rot, aggregators auto-generate pages with no real deal behind them, and the actual announcements live on Instagram and Facebook. The job is not to collect claims — it is to date them.
A restaurant that discounts bottles of wine by roughly half on a recurring night. We search within about 10 driving miles of Hyde Park Square (45208) and list what we verify anywhere inside 20. Glass specials, 25%-off nights and $5-off deals don't qualify, though a few are named on the board so you know they were seen. A weekday afternoon happy hour that meets the bottle test gets its own column rather than being folded in with the evenings. Distances are driving miles from Hyde Park Square by the fastest Google Maps route on the day we checked, rounded to the half mile. A place between 10 and 20 miles that we come across, or that a reader sends us, gets exactly the same check and is listed with its distance, marked “beyond our usual radius”. We don't search that far out systematically, so that part of the list is not complete. Past 20 miles we don't list.
Every listing carries one of four markers. Direct check — we phoned the restaurant or went in and asked, within the last six months — or the last year where the deal has a track record: the same night confirmed or announced by the restaurant six months or more apart, or an end date the restaurant gave us. It outranks the other three because it is current and first-hand; it is the one tier without a link, so the place page gives the date instead. means the deal is printed on the restaurant's own specials page or deal listing. Their own post means the restaurant announced it on its own Instagram or Facebook within the last year. Secondhand means we only have reviews or local press — call ahead. In print the four are a square with a filled centre, a filled, a half-filled and a hollow square so they survive a black-and-white printer.
A silent homepage is not evidence. For each candidate we pull the site's sitemap and read every page on a small site, and on a larger one the specials, events, happy-hour, nightly, weekly, deals, menu, drinks, hours and location pages. Then we search for the restaurant's own social posts, one restaurant at a time, because that is where deals get announced and quietly withdrawn. We check the city on every hit — restaurant names repeat across the country, and a same-name place in another state has fooled us more than once. A dated post, review or article is good for about six months from its date — a year where the restaurant has announced the same night across six months or more — and after that it makes a deal unconfirmed, not confirmed, however long its link keeps working. Listings whose proof is the restaurant's own post are re-read in a browser once a month; a newer post refreshes the date, a post that says the deal changed or ended is acted on at the next monthly check. A specials page that is still live on the restaurant's own site counts as current on the day we read it, whatever its last edit — which is why we re-read every one on each full check.
The board's checked section names every place we looked at, with the date of the last look, and separates the ones whose full site we read from the ones we could only check by search because their site refused us. A search-only check is never described as cleared.
A monthly pass re-runs the sweep on every listing and hunts for new ones: it searches the phrases restaurants actually use — “half price wine”, “wine Wednesday”, “½ price bottles” and the like — crossed with every neighbourhood inside and just past the radius and with Instagram, Facebook and local listing sites, then asks the plain question — who has a half-price wine bottle night near Hyde Park? — of its own search; now and then a person puts the same question to other AI search tools and brings the names back to be checked. Every name that turns up is a lead, screened the same way as a submission, never taken on trust.
No sweep finds everything. A restaurant can start a wine night the day after our pass, and some never put their specials anywhere we can read. If you know one we're missing, tell us and we'll check it.
Every Monday a check works through the submissions we've received. Pages our monthly check can't read as text — image-only specials sheets, sites that block automated reading — are opened in a real browser on our own computer once a month and filed from what they actually show. Each listing shows the date it was last confirmed. Every listing links to the page or post that proves it — the restaurant's own specials page, its own post, or a menu — or says it was confirmed by phone. A listing nobody can confirm gets one more monthly check and then comes off the board. A restaurant's own newer word that a deal has ended removes it at once, whatever older evidence said. The whole dataset is public as JSON.
Questions, corrections, or a restaurant that wants its listing changed: write to hello@halfpricewinecincinnati.com, or use the submission form. A link to the restaurant's own post or specials page is the fastest way to get something changed.
We read each of these restaurants' own websites in full — sitemap, specials, events, menus — and found no bottle deal of about half off. Every name is re-checked at least every 90 days; the date shows the last look.
These have no website to read, or the site could not be reached or read as text, so they were checked only by web search and social posts. They are unconfirmed, not cleared, and are re-checked on every monthly pass; pages we could not read as text are opened in a real browser by the monthly sweep.