An AI answer cited your brand. It feels like progress. Then a competitor shows up more often, a product detail is wrong, and you are not sure which fire to fight first.
Constance Tan, Product Marketer at Ahrefs, tackled that exact problem in a Search Engine Journal webinar. Her starting point: stop treating a single reassuring mention as visibility. Look for recurring patterns across the customer journey, then choose the response that fits.
Build prompts around how customers actually choose
Tan recommends organizing AI monitoring around three question types instead of firing hundreds of random prompts:
- Problems customers need to solve. Capture discovery before a buyer has chosen a brand.
- Positioning and comparisons. See which brands AI recommends for specific audiences, use cases and categories.
- Facts about your business. Test whether answers describe pricing, capabilities, availability and product use correctly.
Use language from search queries, support conversations, sales questions and community discussions. The most useful forums vary by market; Reddit and Quora are not the answer everywhere.
Turn competitor mentions into a source map
If a rival appears more often, treat it as an AI share-of-voice gap. Read the actual answers. Does AI repeatedly favor another brand for small business or ecommerce? Which pages supply those distinctions, and does your content explain the same use cases?
Tan suggests looking beyond articles: reviews, discussions and videos can shape what an answer cites. For outreach, prioritize pages cited repeatedly and authors you can realistically reach. Domain strength and organic traffic add context, but a strong competitor-owned page may leave little room for correction.
Fix what you own before chasing what you do not
Before commissioning another article, audit your own pricing pages, product explanations, profiles and older posts. Conflicting information on your site can leave AI answers describing features or plans that have changed.
Third-party corrections require patience. Ahrefs contacted 26 authors about inaccurate information: 10 replied, and four updated their content. One issue involved outdated descriptions of which plans included API access.
Tan is careful about conclusions: those are outreach outcomes, not proof of a rise in citations or revenue. That is why she selects reachable, frequently cited sources.
When outreach is not practical, her alternative is direct:
Sometimes outreach is not always the answer. Sometimes it’s better to create the new sources of information, new pages that answer or cover the topic in a better way, a more comprehensive way, or with more up-to-date information.
Match the fix to the failure
- Build a balanced prompt set across customer problems, comparisons and factual questions.
- Investigate repeated claims and the sources behind them; ignore every isolated odd answer.
- Correct owned information first, then pursue third-party corrections you can realistically secure.
- Check bot access: firewall restrictions, broken URLs, timeouts and JavaScript-heavy pages can block retrieval.
- Review visibility alongside conversions, sales and customer attribution rather than treating a citation as a sale.
In communities, answer technical questions and correct factual mistakes. Recurring complaints may reveal product or onboarding problems worth taking back to the teams that can fix them.
Start with one customer segment and monitor that prompt set over time. The recurring sources, positioning claims and factual errors give you concrete work. A lone citation does not.
Source: Search Engine Journal
