ExplainerAI visibility
An AI Engine Is Quoting Your Old Pricing. Here Is How to Fix It.
by the AEO GEO Labs team3 min read
One of the most common findings in an AI visibility audit is not absence but error: the engine names you, then quotes a price you changed two years ago, or a plan you discontinued, or an integration you never had. A wrong answer about you is often worse than no answer. It sends the buyer to a competitor whose facts the engine got right.
Why it happens
Engines compose answers from several sources, and for facts about a company they often prefer independent ones. If a software directory lists your 2023 pricing, a review platform profile shows an old plan name, and a years-old comparison article has your previous feature set, the engine may weight that consensus over your own current pricing page, especially if your page is hard to parse or blocked to its crawler.
Sometimes the stale fact is on your own site: a forgotten landing page, a PDF, a help article, or schema that was never updated. Engines do not know which of your pages is canonical unless you tell them.
Finding the source
- Reproduce the error. Ask the engine the question that produced the wrong fact. Note the exact wording and any citations.
- Read the citations. Perplexity and ChatGPT search mode link their sources. Often the wrong fact is right there on a cited page.
- Search the wrong fact. Put the old price or the discontinued plan name in quotation marks and search. Every page that surfaces is a candidate source.
- Check your own site. Site-search the old value. Check schema with a validator. Check llms.txt if you have one. Check downloadable PDFs.
- Check the profiles. G2, Capterra, Crunchbase, Product Hunt, app marketplaces, Wikidata, and any directory in your category.
In most cases this turns up two to six sources carrying the wrong fact. That is the correction list.
Correcting each type of source
Your own pages. Update, redirect, or remove. Add a visible "updated" date to the canonical pricing page. Make sure the price is in the initial HTML, not rendered later by a script.
Your schema. Align Product or SoftwareApplication offers with the visible price. Validate.
Review platforms and directories. Most have a vendor claim-and-edit flow. Claim the profile, correct it, and keep a reminder to recheck quarterly.
Third-party articles. Email the author or editor with the correction and a link to the current page. Many publications update; some add a note. A polite, specific request works far better than a legal tone.
Community threads. A disclosed, factual reply in the thread ("I work at X; pricing changed in March, current plans are here") is both honest and citable.
Wikidata. If you have an item, edit it with a source. If you do not, creating one with correct, referenced facts gives engines a neutral anchor.
How long correction takes
Engines re-read sources on their own schedules. Your own pages typically update in days. Review profiles within a week or two of the edit. Third-party articles depend on the editor. Training-based answers, where the model is not searching live, lag longest and sometimes persist for months, which is why the search-mode corrections matter most: they are what the engine reads today.
Preventing the next one
Keep a single source of truth for the facts engines quote, pricing, plans, integrations, founding date, headquarters, and push changes outward from it: site, schema, llms.txt, profiles. Recheck the question set that produced the error after every pricing change. AEO GEO Labs builds that recheck into weekly reporting precisely because the error usually returns from a source nobody remembered.
If this is happening to you now
Reproduce it, read the citations, and search the wrong fact in quotes. You will likely find the source in ten minutes. Correcting it across every profile and article is slower, which is work AEO GEO Labs does routinely; you can see which engines are currently wrong about you in the free report.