Not sure where to start? Take a guided tour.
AI Site Grader
AI Query Scanner
llms.txt GeneratorThe short answer
AI models often rely on stale training data or cached web crawls that prioritize old product pages with high historical SEO authority. To fix this, you must implement aggressive technical redirects and update structured data to explicitly mark items as discontinued.
- Quickest fix
- Implement 301 redirects from discontinued product URLs to their direct successors.
- Most common cause
- High authority legacy pages remaining live or returning 404s instead of 301 redirects.
Diagnosis
Symptoms
- AI chatbots recommend products you no longer sell.
- Search engine AI overviews cite legacy spec sheets.
- Customers mention ‘finding’ products online that are out of stock permanently.
- Perplexity or ChatGPT cite your old documentation as ‘current’ evidence.
How to confirm it
- Prompt ChatGPT or Claude: ‘What is the best [Brand Name] product for [Use Case]?’
- Check if the AI provides a link to a discontinued page.
- Inspect the ‘Sources’ cited by Perplexity or Google Search Generative Experience (SGE).
- Verify if the cited pages still exist or are returning a 404 error.
- Severity
- Medium
- Urgency
- Moderate. It causes customer frustration and lost conversion opportunities.
- Impact
- Reduced ROAS, increased customer support tickets, and brand perception of being ‘out of date’.
Common causes
| Cause | How to identify it | How common | Effort |
|---|---|---|---|
| Stale training data | The AI mentions the product but cannot provide a live link, or the link is from a 2021-2023 archive. | Very common | Hard fix |
| Zombie pages | Old product pages are still ‘Live’ but marked as ‘Out of Stock’ instead of ‘Discontinued’. | Common | Easy fix |
| High historical backlink authority | The discontinued page has more backlinks than the new version, causing RAG systems to prioritize it. | Common | Medium fix |
| Missing schema markups | Product schema still shows ‘InStock’ or lacks the ‘discontinued’ itemCondition property. | Sometimes | Easy fix |
| Third-party retailer listings | AI is pulling data from Amazon or eBay listings of ‘Renewed’ or ‘Used’ versions of your old product. | Sometimes | Hard fix |
Quick wins
Add ‘[Discontinued]’ to the HTML Title tag of the old product page.
AI RAG systems will see the status in the most prominent metadata.. Time: 5 minutes
Update your brand’s official ‘Current Lineup’ page.
Provides a fresh source of truth for AI to crawl.. Time: 1 hour
Request a ‘Clear Cache’ via Google Search Console for the specific URL.
Forces AI crawlers to see the ‘Out of Stock’ or ‘Redirect’ status faster.. Time: 10 minutes
Solutions
Implement 301 redirects to successor products
Timeline: Immediate implementation; 1–2 weeks for crawlers
Map legacy URLs
Identify every discontinued product URL currently indexed.
Tools: Google Search Console, Screaming Frog
Apply 301 Redirects
Redirect the old URL to the specific replacement model, not the homepage.
Tools: HTACCESS, Redirect Plugin
Update schema.org ItemCondition
Timeline: 1 week
Modify JSON-LD
Update the ‘availability’ to ‘https://schema.org/Discontinued’ or ‘EndOfLife’.
Tools: Schema Generator
Add ‘isRelatedTo’ property
Use schema to point the AI toward the newer model as a replacement.
Tools: Code Editor
Deploy ‘Product sunset’ landing pages
Timeline: 3 days
Create a Legacy Hub
Keep the page live but add a massive header: ‘This product is discontinued. See the new version here.’
Tools: CMS
Update Metadata
Change Title Tag to include ‘[DISCONTINUED]’ to signal AI crawlers.
Tools: SEO Plugin
Refresh external knowledge bases
Timeline: 2–4 weeks
Update Wikipedia and Wikidata
LLMs heavily rely on Wikidata. Ensure your product entries reflect ‘End of Production’ dates.
Tools: Wikidata.org
Submit updated Sitemap
Force a recrawl of your site to show the removal or change of status.
Tools: Google Search Console
Use noindex tags on legacy documentation
Timeline: 1 week
Identify PDF manuals
AI often finds discontinued products via old PDF manuals hosted on your server.
Tools: File Manager
Apply X-Robots-Tag
Add ‘noindex’ to the header of legacy PDF files.
Tools: Server Config
Aggressive internal link pruning
Timeline: 2 weeks
Audit Footer and Navigation
Ensure no site-wide links still point to discontinued products.
Tools: Link Auditor
Update Blog Content
Edit old blog posts that mention the product to include ‘Update: This has been replaced by...’ links.
Tools: CMS
Example scenarios
These scenarios show how the fix plays out. They are illustrative, not customer results.
- Situation
- A high-end audio brand found ChatGPT recommending a 2018 headphone model that had zero inventory.
- Diagnosis
- The old product page had 500+ backlinks, making it the ‘most authoritative’ page on the site according to AI crawlers.
- Fix
- Implemented a 301 redirect and updated the Wikidata entry for the product series.
- Result
- AI mentions shifted to the 2024 model within 18 days.
- Lesson
- Backlink authority can override current inventory status if not redirected.
- Situation
- SaaS company’s ‘Legacy Plan’ was being cited as the primary pricing option by Perplexity.
- Diagnosis
- The legacy pricing PDF was still hosted and publicly accessible, appearing in the top search results.
- Fix
- Applied noindex to the PDF and created a ‘Plan Comparison’ page that explicitly labeled the old plan as ‘Retired’.
- Result
- AI correctly identified current pricing in 90% of test queries.
- Lesson
- Hidden PDFs are a major source of AI misinformation.
- Situation
- Appliance manufacturer saw AI recommending a recalled model.
- Diagnosis
- The recall notice was less prominent than the original product marketing page.
- Fix
- Used a ‘Warning’ schema markup and updated the page title to ‘RECALLED / DISCONTINUED’.
- Result
- AI immediately began citing the recall notice instead of recommending the purchase.
- Lesson
- Urgency signals like ‘RECALLED’ are picked up quickly by safety-aligned AI.
Prevention
Strategies
- Create a ‘Sunset Protocol’ for every product launch that includes URL redirection.
- Maintain an ‘Official Product Registry’ page that AI can use as a primary source.
- Use versioned URLs (e.g., /v2/) to make it clear which is the newest iteration.
Monitoring setup
- Set up a monthly ‘AI Audit’ prompt to check for mentions of retired SKUs.
- Monitor Google Search Console for clicks on ‘Discontinued’ URLs.
Warning signals
- Increase in customer service queries for ‘Where can I buy [Old Product]?’
- Old URLs appearing in ‘Top Pages’ reports in Analytics.
When to escalate
- Signs
- AI is hallucinating that a discontinued product is ‘New for 2025’.
- Redirects are being ignored by major LLMs for over two months.
- Options
- Submit a formal ‘Inaccurate Information’ report to OpenAI/Google.
- Consult an AI Engine Optimization (AEO) specialist.
- Timeline
- Escalate if 30 days of technical fixes yield no change in AI responses.
Frequently asked questions
AI models don’t just ‘see’ your live site; they remember it from previous training runs. If you delete a page (404), the AI might rely on its ‘memory’ of that page. Using a 301 redirect is better because it actively tells the crawler that the old information has moved to a new, specific location, forcing a memory update.