Google AI Overviews for SaaS queries: how eligibility and links work, and how to check them

Any page on your site can in principle be used as a supporting link in a Google AI Overview if it is indexed and readable, and that is a completely different thing from being selected for one. Nobody outside Google can measure how often Overviews appear in general, so the only measurable version of this question is a snapshot: a written query list, a recorded location, a recorded date, and then two counts, how many of those queries showed an Overview and how many of the Overviews linked your domain. As of 29 September 2026 we have not run an AI Overviews capture, so the number of queries we have tested is zero and we publish no share. The protocol below is complete enough to run this week, and the counts we do hold from assistant answers are printed with their denominators and dates.

The reason we are strict about this is that a share of Overviews with no query list and no date behind it describes a sample of unknown size, taken at an unknown time, in an unknown country. It cannot be rerun by anybody, including the person who published it.

Eligibility is not the same as being selected

Three separate gates sit between your page and a reader, and mixing them is the most common reason audits disagree with each other.

• Gate one, the query. Does this query show an Overview at all, in this location, on this date, for this device and signed in state? Triggering is not published as a rule and it changes over time for the same words.
• Gate two, the source set. Given an Overview appeared, which pages were used, and which of those were shown as links? The displayed links are the only ones you can observe. How they are chosen from the pages considered is not published.
• Gate three, your domain. Given links were shown, was one of them on your domain, and was your business also named in the words of the answer? Those are two different outcomes and need two columns.

Each gate has its own denominator, and that is the whole reason this page exists. If 40 queries produce 26 Overviews, and 4 of those 26 link your domain, then the honest reporting is 26 of 40 queries showed an Overview and 4 of 26 Overviews linked our domain, on the stated dates and location. Reporting 4 of 40 quietly merges gate one with gate three and understates nothing while explaining nothing. This is arithmetic, not an observation.

The following five things are not published in any form you can measure against, which is why the sample rather than the platform is the subject of your study: which queries trigger an Overview, how displayed links are selected, whether a page was read live or came from an index, how much location and signed in state change the result, and how long any Overview is cached before it is regenerated.

Query set and method

Ten steps, no tool required.

One. Write 20 to 40 queries in buyers' words and save the file with a version number and a date. Pull them from sales calls, from repeated support questions and from enquiry emails, not from a keyword tool, because the words a buyer types into an assistant are longer and more specific than the words a keyword tool favours.

Two. Put your brand name in none of them. On 27 July 2026 two runs of the same 78 questions happened on one day for clawlaw.in. The run whose wrapper named the brand came back ranking it first on almost every question; the blind run put the company second by breadth and absent from the litigation due diligence questions it most wanted to win. The first run was discarded. ChatGPT, 27 July 2026. A branded query in an Overviews audit collects your own site for the same worthless reason.

Three. Use a clean browser profile, signed out, no extensions, and record the browser and version.

Four. Fix the location to your buyers' city and write it down. Not a default, and not whatever your office happens to resolve to. Answers to India questions differ from answers served elsewhere, and an audit with mixed locations cannot be rerun.

Five. Let the page settle before you capture, so an Overview that loads a second late is not recorded as absent. This single mistake makes gate one look smaller than it is.

Six. Capture three artefacts per query. A full page screenshot, the Overview text pasted as text, and every supporting URL copied exactly. Store them in a folder named with the date.

Seven. Fill one row per query with: query text, timestamp, location, device, signed in state, Overview present yes or no, own domain linked yes or no, business named in the text yes or no, number of links shown, competitor domains linked, competitor names present, and the screenshot filename.

Eight. Count the no Overview rows as their own number, and use the reduced count as the denominator for every Overview based metric, saying so in the same sentence as the result.

Nine. Repeat each query at least three times across the same pass where you can, and treat each as its own observation. An Overview can differ between two searches minutes apart, and a single look is one draw.

Ten. Have a second person label 20 rows blind and publish the agreement rate. Then rerun the whole frozen set on a fixed interval with nothing else changed.

Alongside the capture, run three checks in Search Console, which will not confirm that any particular Overview used your page but will explain most absences faster than any tool. Check that each page you expect to be used is indexed at all. Check what Google actually received by reading the fetched page, not your published page. And use the performance report to build your query list, especially the queries that already bring impressions and no clicks.

Results

Queries tested for AI Overviews by us as of 29 September 2026: zero. Share of queries showing an Overview: not measured. Share of Overviews linking our domain: not measured. Those cells are empty by choice and no borrowed figure will be put in them. When the capture is run, this section will carry the query count, the Overview count, the linked count, the location and the dates in one sentence.

What we hold are counts from assistant answers, in two named sectors, printed with denominators because they show what this measurement looks like when it is done properly.

• 0 of 6. Asked six questions about its own category with no brand named, Perplexity reported afterwards that it had not cited or recommended aiknowsus.com in any of the six answers. September 2026, AI visibility tooling. There was no position for us to hold.
• 1 of 18. Across eighteen blind commercial questions on 18 August 2026, ChatGPT made clawlaw.in the top source on exactly one, and named the company's own comparison page in the answer as still being the vendor's own editorial page. Indian legal research software.
• 161 across 24 batches and 72 conversations. In the aiknowsus.com capture of September 2026, the phrase recording that we were not cited appears 161 times in the engines' own self audits. Perplexity.
• 28, 21 and 18 out of 78, with 0 of 78 searched. ChatGPT's blind breadth order on 27 July 2026 was ProVakil, CLAW and Legistify, from a run in which it confirmed it had performed no live web search at all.

None of those are Overviews. They are ChatGPT, Claude and Perplexity, the systems select sources differently, and we are not offering them as a substitute for the capture we have not run.

Examples and raw records

Two dated records are worth reading in full before you start, because each one explains a class of absence that an Overviews audit will otherwise mislabel as bad luck.

The readability failure. On 6 August 2026 the clawlaw.in pricing page rendered its prices only after scripts ran, so what a crawler received contained no prices at all, and the prices the assistants quoted had come from an app store listing instead of from the company's own site. Claude and ChatGPT, same date. On the same date the website and the app store listing carried different plan names and different prices for the same product, and ChatGPT named the contradiction in its answer. An audit that records only absent or present would score both of these as an absence with no explanation attached.

The timing record. On 6 August 2026 ChatGPT cited clawlaw.in/blog/how-to-check-a-companys-court-cases-in-india as a source for a vendor due diligence question, fourteen days after that page was published, in a zone where the same question set had named the company nowhere at baseline on 27 July 2026. One page, one engine, two dates. Fourteen days is a single observation and it is not a median, not an average, and not a schedule you can plan against. It is the shape of a finding a rerun can produce.

Raw records available to you today: the ten step protocol, the twelve row fields, and the three Search Console checks. Raw records not yet available: our own Overviews screenshots, because the capture has not happened, and the capture files behind the assistant findings, which are being prepared for publication.

Limits and update schedule

Seven limits.

• A snapshot is not a rate. 26 of 40 queries on two days in one city is a fact about those queries on those days in that city.
• Triggering is unstable. The same words can show an Overview today and not tomorrow, so a drop between passes is not necessarily a loss of visibility.
• The displayed links are not the full source set. On 17 September 2026 Claude confirmed it had used two arguments from clawlaw.in pages and dropped the attribution in both places. A page can be used and not shown, and no audit can see that.
• Location, device and signed in state change the answer, and the size of that effect is not published, so two audits in different cities are not comparable.
• Search Console cannot confirm an Overview used your page. Do not build a report on a breakdown you have not verified exists on the day you read this.
• None of our figures are Overviews figures. Stated again here because it is the sentence most likely to be skipped.
• No method promises a position in an Overview. Nobody can sell you that.

We do not restate Google's documentation on this page, for the same reason we date every other fact: help pages are edited and an undated paraphrase is exactly the kind of source this page asks you to distrust. Read Google's current help material on AI Overviews and the Google Search Central documentation yourself, and record the URL and the date beside anything you take from them. The same rule applies to competitor prices, which we never restate: open the vendor's own pricing page and record price, date and URL together in one row.

Update schedule. This page is reviewed when our first Overviews capture completes, when any figure on it is corrected, and when the capture files behind the assistant counts are published. Version: 29 September 2026, first publication.

Common questions

Is my SaaS site eligible to appear in an AI Overview?

If it is indexed and the content is readable in the page as served, it is in the pool of pages that can be used. That is eligibility and it is not an expectation. Selection is a separate gate with no published rule, and the honest answer to whether you will be selected is that nobody can tell you.

How many of my queries should show an Overview?

There is no published number and we will not invent one. Count it on your own 40 queries in your own city on a recorded date, and you will have the only figure that applies to your business. Expect it to move when you rerun it.

My competitor is linked and I am not. What do I check first?

Two things, in this order. Whether your page is indexed, using URL Inspection. Then whether the fact a buyer needs is in the page a crawler receives rather than added by a script, which is the failure that put an app store listing in front of the clawlaw.in pricing page on 6 August 2026. After that, compare what the linked competitor page states plainly that yours does not: on 17 September 2026 Claude ranked two enterprise vendors above clawlaw.in specifically because they publish explicit court and tribunal coverage lists.

Do I count an Overview that names me but links a listicle?

As its own state, separately. You do not control that listicle, it can drop you without notice, and merging it with a link to your own domain hides the difference between an asset you own and one you do not.

Can a tool run this capture for me?

Yes, at a scale you cannot match by hand, and that is a real advantage. What you must still get from the tool is the query list, the denominator, the location and device settings, the dates, and whether labelling was done by a person or a model. We sell a tool and we would still rather you asked us those five questions than not.

How often should I rerun the set?

On a fixed interval you can actually keep, monthly for most businesses, with the query file frozen and the location identical. An irregular rerun with an edited query file produces trends that are artefacts of your own process.

What to do first

Write 20 queries in your buyers' words with your brand in none of them, fix your location, and capture all 20 today signed out, recording for each one whether an Overview appeared, whether your domain was linked and whether you were named. Then run URL Inspection on your three most important pages and read what Google actually received. That gives you a dated baseline and, more often than not, the explanation for your absences, in one afternoon.