How do you stop AI citing old prices after a social-media update?

Agree the authoritative version and effective time

Before publishing, record the product or class, old and new value, start and end times, applicable channels, exclusions and approver. 'This week's deal' leaves expiry, sizes and store-versus-online scope uncertain. Use full dates and time zones for time-sensitive terms. Say whether tax, shipping or materials are included. An undecided change remains unapproved; neither a post nor AI should present it as a commitment.

In a fictional pottery studio, a class was 800 and a new-customer weekend trial is 650, only for specified sessions with group seats checked by staff. Put those conditions together. A social image saying 'from 650' needs a caption with validity and limits; the class page supplies the full rule and enquiry route. AI may cite approved terms, but cannot infer universal eligibility from the image or treat a staff slot as group capacity.

Google's SEO and people-first English guidance recommends useful, maintained web content. Social posts reach followers, while a readable authoritative page explains terms to visitors arriving from search. Google Business Profile also separates regular and special hours, illustrating that public entry points need their own updates; it does not synchronize kakinang knowledge for the owner.

Update every affected source and activate knowledge

Inventory product prices, service descriptions, FAQs, home promotions, old enrolment links, image text, translations and external business profiles. Update or retire each. Removing a social post alone leaves search visitors finding an old campaign page. A useful archived explanation may remain if it clearly says ended and points to current terms.

kakinang site knowledge has candidate content and an activation step. After editing a page, verify source reachability, successful synchronization, candidate text with the new price, activation by an authorized manager and a real AI question. Update price text in FAQs and supported files too. Formal product options should provide current prices. A social post is not guaranteed to be synchronized, and words inside an image are not authoritative product records. Until verified, say the price needs checking and send the enquiry to staff.

Use five columns: source, owner, public state, synchronization state and test result. List web page, FAQ, product or service record, external listing and old link for each change. An approver signs off, then an operator synchronizes and tests. If an agency manages social posts but a manager owns the website, assign each action and completion time; 'posted' in a group chat does not mean all entry points agree.

Test stale questions and exceptions

Ask at least 'what is the price now?', 'does yesterday's old offer apply?' and 'does next month's class qualify?' Repeat with translated queries, variants, dates and campaign codes. Answers must distinguish current price, ended offer and future or inapplicable terms. If AI cites an old price, record the source, correct or disable stale knowledge, synchronize and activate, then retest. A prompt saying 'use the latest price' does not repair conflicting sources.

Staff also need the same version. When a customer holds an old screenshot, follow published terms and the studio's handling rules; AI cannot grant an exception or deny a claim on its own. Give staff a card with old announcement date, current rule and questions needing a decision. For postponement, early end or cancellation, address confirmed customers before changing public information.

Weekly, review stale answers, unfinished synchronization, visits from old links and the questions staff corrected most. If an expired page still attracts visitors a year later, fix its status and link. Assign a translation owner if English is repeatedly missed. Define 'published' as public sources and a real AI answer passing review, not merely a social scheduler success. This reduces false promises without guaranteeing rankings, revenue or immediate change across all outside AI.

Change is not limited to price. Hours, delivery promises, staff, venue, cancellation and whether new work is accepted can all be contradicted by old content. Assign risk: price and cancellation require authorized approval, time and location need every visible entry point updated, and decorative copy may follow. A branch-only offer must not appear on another branch's page. Membership eligibility should appear in both post and site, not emerge only at checkout.

Replay last week's actual customer questions across article, FAQ, mobile chat and external listing. Record each answer's URL, source date and reviewer. If the price page changed but the FAQ did not, do not mark the release complete. A successful synchronization job alone does not prove the new candidate is the active version. Login or crawler blocking may prevent a page from entering candidate knowledge; check reachability rather than use a social screenshot as a substitute.

Give expired offers an exit routine: schedule a reviewer before expiry, remove the home card on the date, update product or service options, mark the campaign page ended, review linked ads and messages, then retest AI. Existing formal orders or bookings follow their confirmed terms and shop policy; a page revision cannot erase a prior commitment. A lead that expressed interest without formal confirmation still needs a clear human decision on price and place.

Record what remains unknown. If stacking a new offer with an older coupon is undecided, AI should say the combination needs confirmation and route to staff. Once the owner decides, update FAQ, campaign page and knowledge and test conflict questions. A solo shop can do this with one authoritative sheet and a brief weekly expired-offer check; sophisticated tooling is optional, ownership is not. If staff report that the post and site disagree, log the gap, check sources and effective dates, pause unsupported automated answers and have an authorized owner decide what to say to existing and new customers before retesting.