Last updated: 16 July 2026
You can check this yourself, free. Pick 10 to 15 real buying prompts, run each one in ChatGPT and four other AI engines, and note whether your brand shows up. Score every result from 0 to 3, then work out the share of prompts where you appear. Low score? Fix your content, then re-check monthly.
Ask the AI engine the questions your buyers ask, then read the answer for your brand name. There's no dashboard for this yet, so you do it by hand. Open ChatGPT, type a buying question, and check three things: does it mention you, does it cite you as a source, and does it put you on a shortlist.
That's the whole idea behind our SCR framework. Every answer lands in one of four states:
Recommended is what wins deals. This matters more every quarter: G2 reported in 2026 that half of B2B software buyers now start their research with AI chatbots. If the chatbot doesn't name you, you're not in that first shortlist.
Test the questions a buyer types right before they pick a vendor, not brand-name searches. If someone already knows your name, the engine will find you. What you want to know is whether you show up when a buyer describes their problem and asks for options. Write 10 to 15 of these.
Use these templates and swap in your category, use case, and buyer:
Keep them phrased like a person, not a keyword. The closer they sit to a real buying decision, the more useful the result.
Don't stop at ChatGPT. Buyers use different engines, and each one pulls from different sources, so your visibility can swing hard between them. Check the same five every time. Running one prompt across all five is one row of data, and the gaps between engines tell you where the work is.
| Engine | Why it matters | How to check |
|---|---|---|
| ChatGPT | The default AI for most buyers and the one they trust for shortlists. | Ask your prompt in a fresh chat. Turn on browsing for current answers. |
| Perplexity | Answer engine built on live citations, so it shows exactly which sources it used. | Ask your prompt, then read the numbered sources under the answer. |
| Gemini | Google's assistant, tied into its search index and Workspace users. | Ask in the Gemini app and check for named brands and links. |
| Google AI Overviews | The AI answer at the top of normal Google results — huge reach, no click needed. | Search your prompt on Google and read the AI Overview box. |
| Copilot | Built into Windows, Edge, and Microsoft 365, so it reaches enterprise buyers. | Ask your prompt in Copilot and check the response plus its citations. |
Why bother with all five? Because search behavior is shifting. Gartner forecasts that by 2026, search engine volume will drop 25% as users turn to AI chatbots and virtual agents. The answer box is becoming the front page.
Give every prompt-and-engine result a number from 0 to 3, then average it out. This turns a pile of screenshots into something you can track over time. A spreadsheet is all you need. One row per prompt, one column per engine, and a score in each cell.
Then pull two numbers out of the sheet:
Run the same sheet monthly and you'll see the trend, not just a snapshot. For the weighted version of this, see how we measure the SCR score.
A low score usually means the engines can't understand who you are or can't find you cited anywhere. You fix that with content, not ads. Work through four categories in order, publish the changes, then wait for the engines to re-index before you judge the results.
Then give it time to land. In our experience, re-indexing typically takes about 2 to 7 days on Perplexity, 1 to 3 weeks on ChatGPT, and 4 to 8 weeks on Google AI Overviews. Don't re-score the day after you ship.
Re-run the full sheet once a month. That's frequent enough to catch movement and slow enough to let your fixes and the re-indexing take effect. If you've just shipped a big content push, add a lighter check two to three weeks later on the prompts you were targeting, so you can see early signal.
Keep every month's sheet. The trend line across months is what tells you whether your work is paying off, and it's the number worth reporting to your team.
Yes. Every engine on the list has a free tier that's enough for a manual check, and a spreadsheet costs nothing. The only cost is your time. Doing it by hand is also the best way to understand what your buyers actually see.
Start with 10 to 15 real buying prompts. That's enough to spot patterns without turning the check into a full-time job. As you grow, expand the list to cover more use cases and buyer types, and keep the core prompts stable so you can compare month to month.
Because they pull from different sources and update on different schedules. You might be recommended on Perplexity and absent on ChatGPT in the same week. That's normal, and it's exactly why you check all five instead of trusting one.
Not on its own. One mention is a Surfaced signal, worth a point, but the goal is Recommended across a good share of your prompts. Look at your Share of Voice and your Recommended rate over several prompts, not a single lucky answer.
This manual check gives you a real read. If you'd rather have it measured across all five engines, scored against your competitors, and turned into a plan we run for you, that's what we do. Book a 15-minute call