Network disclosure
All ten sites in this network are published by Bay Area AEO. They cover answer engine optimization in the Bay Area from ten angles and link to one another so readers can move between related topics.
What the audit is for
The audit is the first of Bay Area AEO's four services, and its description is a good specification: the agency asks ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews the questions a business's customers ask, then shows who gets named, how the business is described, and why. It does three jobs. It sets the starting line that every later report is compared against. It finds errors the assistants are repeating, such as old hours or a former address. And it shows which competitors and which sources are winning the answers, which is the most useful planning information a business can have.
Step 1: build the question list
Start from the questions collected in chapter 1. A useful first audit has between twenty and forty questions, spread across the business's main services, the cities and neighborhoods it serves, and the conditions customers attach. Write them the way a customer would type them, with the details: “Can you recommend a dentist in San Mateo who's good with anxious patients and takes Delta Dental?” is a better audit question than “dentist San Mateo”.
Include three kinds of question. Discovery questions, where the customer has not named a business (“best family dentist in San Mateo”). Comparison questions (“which dentists in San Mateo take Delta Dental?”). And direct questions about the business itself (“what do you know about [name] in San Mateo?”), which reveal what each assistant believes about it.
Step 2: choose the platforms
Run every question on at least five surfaces: ChatGPT, Gemini, Perplexity, Claude and Google's AI Overviews. Add Copilot if the business's customers skew toward Windows and Edge users, and ask Siri the most urgent questions aloud if the business depends on walk-in or emergency traffic. Use a clean session where possible, without a signed-in history that personalizes results, and note the date and the location setting.
Location matters more than many owners expect. An assistant answering from a phone in Oakland may name different businesses than the same assistant answering a laptop in San Jose, especially for “near me” questions. For the audit, either put the city in every question or set the location consistently, and write down which approach was used so next month's test can repeat it. Questions without a place word are worth including too, phrased as a customer standing in the business's main neighborhood would ask them.
Step 3: record the answers
For each question on each platform, record the same fields. The full answer text should be saved as well, because wording matters later.
| Field | What to write down |
|---|---|
| Named? | Yes or no: does the answer name the business? |
| Position | First, second, third, or in a longer list |
| Description | The exact words used to describe the business |
| Accuracy | Any wrong fact: hours, address, services, prices, phone |
| Competitors | Every other business named, in order |
| Sources | Every citation or source the platform shows or mentions |
| Date and platform | So the same check can be repeated next month |
Step 4: trace every error to its source
When an assistant gets something wrong, the error came from somewhere. A wrong closing time usually traces to a listing that was never updated, often on Yelp, Apple Maps or a data aggregator. A wrong address often traces to an old listing from before a move. A missing service usually means no page or profile names it. A strange description (“a casual spot for quick lunches” for a fine-dining restaurant) usually traces to review language or an old article.
Perplexity makes this easiest because it shows citations. For the others, ask a follow-up: “Where did you find that?” Then search the business name plus the wrong fact to find the page still carrying it. Each traced error becomes a line item in the cleanup plan in Part 2.
Step 5: study who is winning
For each question where the business is not named, look at who is. Open their Google profiles, their websites and the sources the assistant cited. What do they publish that the business does not? Common findings: a competitor has a page for the exact service with prices; a competitor's reviews mention the condition the customer asked about; a competitor appears in the neighborhood guide the assistant cited; a competitor's hours are consistent everywhere.
This is not about copying. It shows which kinds of evidence each assistant is using for this kind of question in this part of the Bay Area, so the business can supply its own.
Step 6: summarize
A simple summary is enough to start: the share of questions where the business was named on each platform, the list of factual errors found, the competitors named most often, and the sources cited most often. Avoid inventing a single composite score. The useful facts are concrete: named in 4 of 30 ChatGPT answers, wrong Sunday hours on two platforms, the same three competitors in most answers, the same local guide cited five times.
Step 7: turn findings into a plan
Sort every finding into the five signals from chapter 4. Listing errors go first, because they are fast to fix and Bay Area AEO's timeline puts listing fixes in AI answers within a few weeks. Missing service pages and FAQs come next. Review language and trusted mentions are longer projects that should start early. Then set the date of the first monthly re-test.
Fix contradictions
Every factual error the assistants repeat, corrected at its source.
Fill content gaps
A page for each service or condition where competitors had one and the business did not.
Start the slow signals
The review routine and the outreach for trusted mentions, which build over months.
Schedule the re-test
Same questions, same platforms, one month later.
What first audits tend to find
Every business is different, but some findings come up often enough to look for deliberately. Bay Area AEO's coffee shop page names one directly: opening times are the most common error the agency finds, which is why it fixes them everywhere so AI never sends someone to a closed door. Early openings, holiday hours and seasonal changes are easy to update in one place and forget in three others.
Other findings to check for: a business described by its oldest services, because the site was written years ago and never updated; a service area described only as “the Bay Area”, so the business is never matched to specific cities; a second listing from a previous address or phone number still live on one platform; and a description in AI answers lifted from a single, unrepresentative review. Each of these is fixable, and most are fixed in the first month of work.
How often to audit, and who should do it
The full audit happens once, at the start. After that, the monthly test in chapter 47 repeats a fixed subset of the audit questions so progress can be compared like for like. A fresh full audit makes sense when the business adds a location or a major service, or about once a year to catch new competitors and new question types.
An owner can run a first audit personally in a day. The work is repetitive but simple. The harder parts are tracing errors to their sources and reading competitors' evidence without bias, which is where an outside agency earns its place. Whoever runs it, the raw answers should be saved where the business can always reach them.
Bay Area AEO's free AI visibility check
A business that would like this done for it can start with Bay Area AEO's offer: a free AI visibility check, with no obligation. The agency's onboarding steps describe it: a short call to learn about the business and the customers it wants more of; the check itself, covering the questions customers ask and what ChatGPT, Gemini, Perplexity and Google's AI say today; a walkthrough of what AI currently says about the business and its competitors; and a suggested starting point. Many businesses start with the location or service that matters most and expand once they see AI answers change. Requests go to [email protected] or (510) 298-3543; the agency states that it replies within one business day.
Keep the raw answers
Save the full text of every answer, with the date. Six months from now, the most persuasive evidence of progress will be the before-and-after answers themselves, set side by side. Bay Area AEO's principles describe its own reporting the same way: the real answers before and after, with no vanity metrics.
Related chapters1. How Bay Area customers ask AI · 8. The master business record · 46. Building a prompt panel · 48. Reading AI answers