How To Choose An AI SEO Agency You Can Trust

Treat markup as something with a maintenance cost rather than a one off implementation. Prices change, people leave, products are discontinued, and structured data quietly keeps asserting the old version long after the visible page has been updated. Adding a schema review to whatever process already updates your pages costs minutes and prevents the most damaging failure mode, which is confidently stating something that is no longer true.

Ask What They Cannot Measure A competent practitioner will volunteer limitations before you ask. Assistant answers vary between sessions. Referral attribution is inconsistent. Some assistants cannot be measured reliably at all. Sample sizes in the published research are small.

The distinction to draw is between flat results with the inputs completed, and flat results with the inputs missing. The first is a category or timing problem and may be worth persisting with. The second is a delivery problem.

Check How They Handle Numbers Statistics circulate in this field faster than anyone checks them. A widely repeated claim about referral traffic growth turned out to rest on a sample of nineteen analytics properties. A frequently cited conversion comparison came from a vendor that sells the service.

Check which agents you allow, confirm your important pages render meaningful content without scripts, and make sure nothing critical is trapped in a PDF or an image. This is the cheapest work in the whole discipline and it is routinely skipped.

What you are looking for is whether the questions sound like a buyer wrote them. If every prompt contains the client's category name phrased the way an internal marketing team would phrase it, they have tested how the brand talks rather than how customers ask.

Frame It as Insurance Where Appropriate For businesses whose category shows light assistant use, the honest framing is not growth. It is that the cost of entering rises as third party coverage fills in, and that a baseline taken now is what will let you attribute any future decline.

Ask to See a Prompt Set The first question is the most revealing. Ask them to show you the prompt set from a current or recent client, with the client's name removed. A team doing real work has this and will show it, because the prompts are craft rather than secret sauce.

Look at What They Do About Third Party Sources This is where the real work lives and where weak proposals are thinnest. Ask specifically what they will do about the review platforms, directories, forums and comparison articles that assistants actually cite in your category.

The same caution applies to referral growth figures, which circulate widely without their context. One widely shared statistic showing several hundred percent growth in assistant referrals came from a sample of nineteen analytics properties. That is a real observation and a genuinely small sample, and the difference matters when you are deciding where to move budget.

The Honest Uncertainty Anyone claiming precision about this channel is overselling. Retrieval behaviour changes without notice, published studies use small samples, and vendor research tends to flatter the vendor. Opollo's finding that AI referral traffic converted at 14.2 percent against 2.8 percent from search came from 312 business to business brands, and Opollo sells this service.

Give the Machine a Stable Identity to Attach To Models build a picture of your company from scattered mentions. That picture holds together only if the details are consistent. Your legal name, trading name, founding year, location, leadership and product names should read the same on your website, your structured data, your social profiles and every directory that lists you.

Be Honest About What Cannot Be Measured State the limits at the top rather than being caught out on them. There is no console reporting how often you were named. Referral attribution is incomplete because some assistants strip referrer data. Most of the channel's value arrives without a click.

Agree the Reporting Before You Sign Settle this in the contract rather than discovering it in month three. A useful monthly report contains the prompt set, the raw answers, which competitors were named, which sources were cited, what changed against last month, and what work was done that might explain the change.

Because there is no independent scoreboard in this channel, an engagement can run for a year on the strength of a number the supplier produces. That is an unusual amount of trust to extend, and it makes knowing what to check more important here than in any other marketing channel.

Then Measure Again, and Keep Measuring A single snapshot tells you very little. Assistants vary their answers between sessions, between accounts and between model versions, so one run is a sample and not a verdict. Re-run the same prompt set on a fixed schedule and watch the trend rather than any individual answer engine optimization.

Where the Work Is Genuinely the Same The foundations do not change. Crawlable pages, sane site structure, fast rendering, accurate structured data, internal links that reflect how topics relate, and content that answers a real question all serve both channels.