Why does an AI assistant quote an old price for your business?
You changed a service from €90 to €120.
Your website now shows €120.
An AI assistant still says €90.
That is not necessarily an “AI visibility” problem. It is a freshness and source problem around one commercial fact.
The useful question is:
Where did the old price come from, and which public source should now be treated as authoritative?
Short answer
An AI answer can repeat an old price for several reasons:
- the cited page itself is outdated;
- a third-party page still shows the old price;
- a search index has not yet processed the updated page;
- the AI answer is using a different source from the one you changed;
- the AI system generated the answer without a sufficiently current source;
- multiple public sources disagree.
OpenAI’s current ChatGPT Search guidance warns that search results and citations can be incomplete, outdated or incorrect.
Google says its AI search features use information from Search and supporting web pages, and that changes may take time to be recrawled and processed.
So the first task is not “make AI use the new price.”
It is:
make the current price unambiguous in the sources you control, identify stale sources, and measure whether the answer changes.
First, capture the exact wrong answer
Before editing anything, record:
- the AI product;
- the exact prompt;
- the quoted price;
- the correct current price;
- date and time;
- citations or source links;
- whether the answer says “from”, “starting at”, “average” or a fixed price.
This prevents a later comparison from turning into memory.
The error should be reproducible.
Open every citation before changing your site
If the AI answer shows sources, inspect them.
Ask:
- Does the cited page still show the old price?
- Is the source your own site?
- Is it a directory, booking platform or archived page?
- Is the quoted figure a historical promotion rather than the current price?
- Did the AI convert “from €90” into “costs €90”?
- Is the citation about the same service or a different package?
A source-backed answer can still be wrong.
OpenAI explicitly recommends checking cited sources because results and citations can be outdated or incorrect.
Make one authoritative current-price page clear
Your own public page should state the current commercial fact plainly.
For example:
Current price: €120. Last updated: 12 September 2026.
Use wording that matches the offer.
Do not publish several competing price statements such as:
- €90 in an old FAQ;
- €120 on the pricing page;
- “from €95” in an old blog post;
- €110 in structured data.
The problem is not that every page must repeat the price.
The problem is that the pages that do state it should not contradict each other.
Should you include the date?
For a changing commercial fact, a visible update date can help people understand freshness.
It can also make later audits easier.
But do not invent a date or automatically refresh a date without a real content update.
The date should reflect an actual change or verification of the price.
What about structured data?
If your page uses structured data that includes price or offer information, it should match the visible page.
Google’s Search documentation says structured data should represent the visible content accurately.
Do not leave an old price in markup while the page shows a new one.
Structured data is not a guarantee that AI answers will use the price.
It is one more place where contradictory data can exist.
What if a third-party site still shows the old price?
That may matter if the AI system is using or citing it.
Prioritise third-party sources that are:
- authoritative;
- frequently visible;
- directly tied to bookings or listings;
- cited by the AI answer;
- clearly presenting the old price as current.
Correct them where you have a legitimate editing route.
Do not create fake profiles or spam update requests.
If you cannot edit the third-party page, document the conflict.
What if the old price appears on your own old article?
Decide whether the article is historical or still presenting the number as current.
If it is historical, label the context clearly.
For example:
Historical price at the time of publication: €90. Current pricing is available here.
If the article is supposed to be current, update it.
Do not silently rewrite historical reporting if the old price is part of the historical record.
The goal is clarity, not erasing history.
How quickly will an AI answer update?
There is no universal guaranteed timing.
Google’s Search Central documentation says crawling and processing changes can take time, from days to longer periods depending on the page and system.
OpenAI does not guarantee that a particular source will be used for a particular answer.
So avoid promises such as:
- “AI will update in 24 hours.”
- “Once Google recrawls the page, all assistants will show the new price.”
- “Adding schema forces the new value.”
Those statements are not supported.
What if the AI answer has no citation?
Then you have less direct evidence about the source.
You can still compare the public information that a search-backed or externally grounded system might encounter:
- official pricing page;
- product/service pages;
- booking pages;
- Business Profile where relevant;
- major directories;
- structured data;
- old press or blog pages;
- cached or indexed public pages.
But do not claim you know which source the AI used unless the product shows it or the platform documents the mechanism.
What if different assistants show different prices?
Record them separately.
One system may show €120 while another still shows €90.
That does not mean one universal “AI database” is half-updated.
Different systems can use different models, search indexes, sources and retrieval methods.
Track each observation by:
- product;
- query;
- date;
- answer;
- citation/source;
- correct price.
What if the AI combines two different offers?
This is common enough to check explicitly.
A business may have:
- setup fee;
- monthly fee;
- audit price;
- “from” price;
- promotional price;
- old package;
- location-specific price.
An AI answer can appear outdated when it has actually mixed two valid but different commercial facts.
Make product/package names clear.
For example:
- Audit: €960
- Initial setup: €1,200
- Monthly care: €1,200/month
That is easier to verify than one paragraph containing several unlabeled numbers.
How do you know the problem is fixed?
Define one checkable statement.
Wrong state
“AI says the current service price is €90.”
Correct state
“The same query now states €120, or clearly identifies €90 as historical rather than current.”
Do not define success as:
- all AI systems always using the price;
- guaranteed citation;
- guaranteed recommendation;
- guaranteed ranking.
Those are outside the business’s control.
What should a business owner check?
- What exact price is wrong?
- Which exact service/package does it refer to?
- What is the current authoritative price?
- Which source did the AI cite?
- Does the official page show the current price clearly?
- Do visible markup and third-party sources agree?
- Are historical prices labelled as historical?
- What exact query will be used for the re-check?
If these are not clear, fix the commercial fact first.
Then measure AI answers.
Discuss AI visibility with Madloba Consult
FAQ
Why would an AI answer show an old price after we updated our website?
The system may be using a different source, an outdated indexed version, a stale third-party page or information that has not yet been recrawled and processed. The citation, if available, is the first place to check.
Does changing the price on our website update every AI assistant?
No. Updating the authoritative page improves the source information you control, but external AI systems choose their own sources and update on their own schedules.
Should structured data contain the current price?
If your structured data includes price or offer information, it should match the visible page. Do not leave contradictory old values in markup.
Should we delete every old price from the web?
No. Historical prices can remain when they are clearly historical. The problem is an old price presented as if it were the current offer.
Can Madloba Consult guarantee when an AI answer changes?
No. We can document the error, correct or clarify sources the business controls and re-test the defined query. The external AI system controls its answer.
How should we verify the correction?
Repeat the same recorded query and check whether the current price is stated correctly or the old price is clearly identified as historical. Record the answer and its sources.