Search AI: AI visibility tracking for ChatGPT, Gemini and Perplexity
Show clients where they appear in sampled AI recommendations. Search AI compares mentions with up to five named competitors and produces a weekly visibility score and suggested next steps under your agency’s brand.
Included in the $99/month plan, with unlimited clients and locations. OpenRouter usage is billed separately.
Written by Mannie Gill, Founder & Lead Developer at EmbedMyReviews. LinkedIn
Search AI queries OpenAI, Google and Perplexity models through your own OpenRouter key. These are model API samples, not recordings of every user’s ChatGPT, Gemini or Perplexity session.
How Search AI tracks AI visibility
Configure the business and provider connection, then weekly scans are scheduled on Mondays. Check completed coverage before comparing results.
- 1
Describe the business
Name, category, town and website, plus up to 5 competitors. Or paste a description and press Generate from Description.
- 2
Six customer questions
One asks about the business by name. The rest ask what a local customer asks, such as "Which dentist do you recommend in Leeds?" Or write up to 5 of your own.
- 3
Repeated model samples
The standard configuration attempts six questions twice across three model providers: 36 answers, with web search enabled. Economy mode uses fewer runs; incomplete responses can reduce the usable sample.
- 4
Checked and scored
Every answer is checked for the business, its competitors and the websites it cites. You get a mention score out of 100 and an action plan.
Every step is in the Search AI setup guide.
A ChatGPT visibility tracker that counts mentions
Search AI summarizes mentions in the sampled answers as a score out of 100. The mention component weighs repeated results by their consistency; failed or unusable responses are not the same as a confirmed absence.
60 or more reads Good AI Visibility. 30 to 59 reads Building AI Visibility. Below 30 reads Low AI Visibility. These are EMR score labels, not a percentage of all AI searches or a Google ranking. Limited-evidence scans have additional score safeguards.
- Up to 60 points
- How often the business is mentioned in usable answers, weighted by consistency.
- 15 points
- When at least one engine knows the business when asked about it by name.
- Up to 20 points
- How many of the three engines mention it.
- 5 points
- When the business's own website shows up in an answer.
Each engine side by side
Mention rates for the sampled OpenAI, Google and Perplexity model answers, so you can compare the monitored providers.
By type of question
How the business does on discovery, recommendation, comparison and review-led questions.
Competitors you name
How often each of up to 5 competitors is mentioned, on which engines, and the gap to the strongest.
The websites AI cites
The sites the answers lean on, by domain: review sites, directories, the business's own site and others.
Change over time
Score change since the last scan, and a trend chart of the last 12 scans.
The answers behind it
Read what each engine said for any result. Answers are kept for 90 days.
Turn AI visibility findings into client work
Completed scans produce recommendations from nine fixed rules. Use those rules to identify a gap to investigate, then check the answers and the client’s circumstances before recommending work.
Not recognised by name
Make the name, category and location match on Google Business Profile, directories and the website.
Rarely recommended in discovery questions
Inspect the cited sources and the client’s actual service coverage. Improve incomplete information and ask customers for honest feedback where appropriate.
Missing on one engine
Check whether the cited directories are relevant and accurate for the business before recommending changes.
Website never cited
Add service pages that answer the questions customers ask.
A competitor mentioned twice as often
Compare review counts and ratings on the sites the answers cite most.
Sometimes mentioned, sometimes not
Check repeated answers and evidence quality first. Keep the prompt and settings stable before attributing a change to new content or reviews.
Several fixes come back to more, fresher reviews. Ask for them with review campaigns.
AI visibility for local businesses
Search AI asks the way a local customer does: the business's category plus its town.
AI answers are one side of local visibility. Google Maps is the other: Local Search Grid shows where each client ranks on Google Maps across their area. For the map side, read our Google Maps SEO guide.
- Typical questions: "Best plumber in Leeds" and "Which plumber do you recommend in Leeds?"
- Online businesses are asked about with "online" in place of a town.
- Each location gets its own scan, and can have its own town and competitors.
- A business with two or more locations also gets one scan for the business as a whole.
- A location list shows which locations need attention first.
Sell AI visibility to every client
Add a clear AI visibility deliverable to your agency service: the prompts monitored, the sampled answers, relevant competitor mentions and the next action. Give clients a weekly trend with enough context to understand what changed.
Your rules, per client plan
Set Search AI to Enabled, Upgrade or Hidden on each client plan. Sell it as an add-on or a premium tier.
Clients see it under your brand
With Search AI on, clients open their own Search AI page in your branded app, on your own domain once you add one.
Email alerts
The business owner can receive alerts for score drops or missing mentions. Minimum sample sizes and evidence-quality checks help avoid treating incomplete scans as definite losses.
Scheduled PDF reports
Add a Search AI report to a client's scheduled reports, weekly, every two weeks or monthly, with your logo and colours.
Runs every Monday
Scans run on their own each week. You can run an extra scan whenever you change something.
Every client, one plan
Run it for every client and location on the $99/month plan.
How to turn a scan into a useful client update
Keep the business, area, prompts and model settings comparable. Open the answers behind a movement before turning the score into a recommendation.
Example agency interpretation
A plumber appears by name but misses a service question
Suppose the sample recognizes the business when asked by name, but does not mention it for “Who offers emergency plumbing in Austin?” That suggests a question to investigate, not proof the business is invisible to all AI users.
- Read the actual responses and confirm the intended business, location and competitors were matched correctly.
- Check whether the client really offers emergency work and whether its hours, service area and website explain that clearly.
- Correct factual gaps in relevant business profiles and useful service content. Avoid inventing availability just to match the prompt.
- Record the changes and compare future scans using the same setup. Treat any improvement as a measured association, not a proven causal result.
Illustrative workflow, not a customer result. Recommendations are starting points for agency judgment, not guaranteed ranking fixes.
Report AI mentions separately from Google Maps rankings, website visits and leads. Use branded Search AI reports to explain the sample and next action; use the client's analytics or CRM to measure actual acquisition.
Runs on your own OpenRouter key
Search AI uses your own OpenRouter API key. OpenRouter bills you directly, and we add nothing on top. Connect OpenRouter as your default AI provider and the weekly scans run on their own.
A standard scan attempts 36 answers: six questions, twice across three providers. Using the model rates checked on 13 October 2026 and an example of 5,000 input and 500 output tokens per answer, budget roughly $0.50 per scan, or $2 for four scans. Actual tokens, web searches, extra checks and economy-mode settings change the bill. Check your OpenRouter usage before setting a client allowance.
Search AI works during the 14-day free trial once your OpenRouter key is connected.
Search AI FAQ
Related features
Local Search Grid
See where each client ranks on Google Maps across their area.
AI Insights
Themes, scores and recommendations from each client's reviews.
Sales Intelligence
Branded prospect reports with Google reviews, map rankings and AI visibility.
White-Label
Your domain, logo and colours across the client portal and reports.
Keep each profile active with Google Business Profile posts, written from each business's confirmed facts and approved by the client from an email link.
See whether EMR fits the way
your agency actually runs.
Try the real workflows, brand the platform, and decide with your own eyes whether it belongs in your stack.