agent's personalization step quoted a competitor's pricing page as

Not specified by the original author.

TL;DR

Language models invent plausible-sounding competitor details. Pull the false claims, verify every competitor fact against the vendor's own page, and require a cited source for any claim the agent writes about a third party.

agent hallucinated competitor pricing and published it to the comparison page

Use this when

• A comparison page states competitor prices you never verified.
• The agent cites features the competitor does not have.
• A competitor complains about false claims on your site.

Not for this skill when

• Your own pricing being wrong, which is a CMS problem.
• Outdated pages from before a competitor changed pricing.
• Opinionated comparisons, which are editorial, not factual errors.

Steps

1. Identify every unverified competitor claim on the page. Verify: you have a list with the exact sentences flagged.
2. Check each claim against the competitor's current pricing or docs page. Verify: every claim has a source URL or is marked false.
3. Remove or correct the false claims immediately; add a correction note if the page was public. Verify: the live page contains no unsourced competitor facts.
4. Add a claim gate: the agent must attach a source URL to every third-party fact it writes. Verify: a test draft without sources is rejected.
5. Schedule quarterly re-verification of competitor claims. Verify: the calendar or cron owns the next check.

Variant phrasings

ai hallucinated competitor pricing

Common phrasing of the problem.

llm invented competitor features comparison page

The features-angle search.

verify ai generated competitor claims

The process-angle search.

Compatibility: Any LLM content pipeline; the failure mode is model behavior, not a specific API.

Why it happens

Models are trained to produce fluent, confident text, and pricing tables are exactly the kind of structured data they confabulate convincingly. An agent with no grounding step will fill the comparison table from pattern memory instead of the vendor's site.

Edge cases / pitfalls

• Even sourced claims go stale; the gate needs a refresh cadence, not just a one-time check.
• Screenshots of competitor pages are evidence for you, not citable sources for readers.
• False pricing claims can draw legal complaints; treat this as a liability issue.
• The agent will hallucinate sources too; verify the URL loads and says what is claimed.

Provenance

Resolved from the public thread: https://vectle.com/posts/pstl49XOLcSq1RJeoAmvAQ

Resolved from the public thread: https://vectle.com/posts/pstl49XOLcSq1RJeoAmvAQ

Thread query: agent hallucinated competitor pricing and published it to the comparison page

TL;DR: Language models invent plausible-sounding competitor details. Pull the false claims, verify every competitor fact against the vendor's own page, and require a cited source for any claim the agent writes about a third party.

Not specified by the original author.

TL;DR

Give the drafting step one approved source for offers and prices, and forbid it from stating any discount, price, or term that does not come from that source. When the source has no matching offer, the agent must say it will check with the team instead of inventing one.

agent's reply-drafting step hallucinated a discount that doesn't exist  -  the lead replied 'great, I'll take the 40 percent off'

Steps

1. Create a single approved-offers file the agent reads at draft time: current discounts, who qualifies, expiry dates, exact wording. Nothing outside this file is a real offer. Expected: the file exists, is dated, and the agent can quote it back verbatim when asked what discounts exist.

2. Rewrite the drafting instruction so any price, discount, or special term must cite a line from the approved-offers file, and if no line matches, the reply says the agent will check with the team. Expected: on a test prompt fishing for a discount, the agent declines to invent one and offers to follow up.

3. Add a review gate: any draft containing a number, percent sign, or words like "discount", "deal", or "pricing" gets flagged for a human or a checker step before sending. Expected: flagged drafts sit in a review queue instead of going out, and the flag log shows why each one was caught.

4. For the lead who already accepted the phantom discount, decide the commercial response up front (honor it once, offer a real alternative, or decline cleanly) and have the AE own the conversation, not the agent. Expected: the lead gets a clear answer from a human within one business day, and the thread is documented.

5. Re-run the reply scenarios that produced the hallucination as regression tests after every prompt change. Expected: all discount-fishing prompts return the safe fallback, and the test log records the pass.

Use this when

• the agent states discounts, prices, or terms nobody approved
• replies commit the company to offers that do not exist
• a lead accepted a phantom offer and now expects it

Not for this skill when

• the pricing was real but outdated - that is a stale-data problem, fix the refresh
• the agent promised a feature that does not exist - see the roadmap-claims skill
• you need help designing the actual discount policy - that is a business decision, not an agent bug

Variant phrasings

• SDR agent invented a 40 percent discount in a reply
• agent quoted pricing that was never approved
• hallucinated offer accepted by the lead
• reply-drafting step making up special terms

Why it happens

Drafting steps are rewarded for being helpful and specific, and a vague "we can work something out" feels less helpful than a concrete number. With no grounded source of truth for offers, the model fills the gap with a plausible-sounding discount - and "40 percent off" is about as plausible as it gets. The fix is removing the gap: every number must come from the file.

Edge cases

• The approved-offers file goes stale after a promo ends - put an owner and an expiry review date on it, or the agent will keep quoting dead offers.
• The checker step itself can miss creative phrasing ("we can be flexible on commercial terms") - review a sample of unflagged drafts weekly.
• Leads will forward the phantom offer internally - assume anything the agent wrote is now a screenshot somewhere, and respond accordingly.
• Different segments get different pricing - the file needs segment rules, or the agent will apply the enterprise discount to everyone.

Provenance

Resolved from the public thread: https://vectle.com/posts/pstkuSllaXI9r8E3TXnB4BaA

Resolved from the public thread: https://vectle.com/posts/pstkuSllaXI9r8E3TXnB4BaA

Thread query: agent's reply-drafting step hallucinated a discount that doesn't exist - the lead replied 'great, I'll take the 40 percent off'

TL;DR: Give the drafting step one approved source for offers and prices, and forbid it from stating any discount, price, or term that does not come from that source. When the source has no matching offer, the agent must say it will check with the team instead of inventing one.

Not specified by the original author.

TL;DR

Do not let the agent answer "how are you different from X" from memory. Give it a sourced battlecard per competitor, require every comparative claim to cite the card, and when there is no card for X, the agent defers to the AE instead of improvising a comparison.

lead asked 'how is this different from X' and the agent's answer described X's features as ours  -  the reply went out before anyone reviewed

Steps

1. Build one battlecard per competitor you actually face: what they do, what you do differently, and what to say when asked - all sourced, dated, and owned by product marketing. Expected: a card exists for each top competitor, and the agent can answer only from the card, not from training data.

2. Rewrite the drafting instruction: comparative claims must cite the battlecard; if no card covers the named competitor, the reply says "great question - I will get you an accurate comparison" and flags the AE. Expected: a test naming an unlisted competitor returns the deferral, never an invented comparison.

3. Add a review gate on competitor names: any draft mentioning a competitor goes to a human or a checker step before sending. Expected: comparison drafts wait in review, and the gate log shows which competitor triggered it.

4. Audit recent sent threads for competitor mentions to find comparisons that already went out wrong, and brief the AEs who own those threads. Expected: a list of affected threads with the exact claims made, so each AE can correct the record on the next touch.

5. Refresh battlecards on a schedule - competitor features change, and a stale card becomes the same lie with better sourcing. Expected: each card carries a last-verified date and an owner who re-confirms it quarterly.

Use this when

• the agent describes a competitor's features as your own (or vice versa)
• differentiation answers go out with no sourced basis
• a competitor name in a draft should trigger a pause

Not for this skill when

• the agent misstates your own features without any competitor involved - that is product grounding
• the agent invents pricing - see the discount-hallucination skill
• you need the battlecards written - that is a product marketing task

Variant phrasings

• agent mixed up our features with the competitor's
• "how are you different from X" answered from memory
• false comparison sent to a prospect
• competitor features attributed to us in a reply

Why it happens

Differentiation questions feel like knowledge questions, and the model has plenty of training data about competitors - most of it outdated, some of it about similarly-named products. It assembles a confident answer from fragments, and the fragments do not respect which company is which. The card removes the assembly step: the answer is retrieved, not composed.

Edge cases

• The prospect names a competitor with a similar name to yours - the deferral path is safer than guessing which one they meant.
• Battlecards can leak if the agent pastes too much - keep cards to talk tracks, not internal strategy docs.
• A prospect corrects the agent's comparison mid-thread - treat the correction as a signal to review the card, not as ground truth to adopt blindly.
• New competitors appear faster than cards get written - the deferral fallback is the default for a reason; use it.

Provenance

Resolved from the public thread: https://vectle.com/posts/pst_gmdBVdgbHJN9Mpv89muV5g

Resolved from the public thread: https://vectle.com/posts/pst_gmdBVdgbHJN9Mpv89muV5g

Thread query: lead asked 'how is this different from X' and the agent's answer described X's features as ours - the reply went out before anyone reviewed

TL;DR: Do not let the agent answer "how are you different from X" from memory. Give it a sourced battlecard per competitor, require every comparative claim to cite the card, and when there is no card for X, the agent defers to the AE instead of improvising a comparison.