Playbook
Exist for AI assistants.
Before an assistant can recommend a company, it has to be able to look the company up. Seven places make that possible. Most of them take a form and twenty minutes. This is the order to do them in, and what each one is for.
Citations are the top layer. Entities are the one under it.
An assistant answers from pages it can cite, and from a smaller set of records that tell it what a company is: a Wikidata item, a Google entity, a review-site listing, a schema block on the home page. When those records agree with each other, the assistant treats six listings as one company. When they are missing, it has nothing to resolve your name against and quietly picks someone it can.
AI Visibility checks these seven for you every week and shows each one as found, missing or confirmed. This page is what to do about a missing one. Use the same name, the same one-line description and the same website everywhere; the sameness is the signal.
1. Wikidata. Do this first, twenty minutes.
Wikidata is the one entity store every assistant reads, and it has no notability review. Make a plain account, search for your company name first, and read any item that already carries it so your description separates you from it. Then create a new item: the label is your name, the description one line that says what you are and where, the aliases your domain and your legal entity.
Add the statements one at a time: instance of (business), official website, country, headquarters, inception, industry, and your LinkedIn company id. Give each a reference where the form asks for one. Save, copy the Q-id, and put its URL in the sameAs list on your home page (step 7). That link is what makes the two ends confirm each other. Keep the description free of praise; editors revert it. Do not write a Wikipedia article about your own company: it will be deleted and the log is public.
2. Google. A Business Profile, then wait.
Google’s Knowledge Graph feeds AI Overviews and AI Mode, and there is no form to submit a Knowledge Panel. Google builds one when the same company is confirmed from several places. Claim a Google Business Profile with exactly your name, your website and your category, verify it by whichever method Google offers, and make sure your home page schema links to your LinkedIn page and your Wikidata item.
Then wait. Search your name and country once a month. The presence check reads Google’s own entity search for the same thing, so the row flips on its own when the graph has you. Keep the profile name plain: Google suspends profiles with a tagline in the name.
3. LinkedIn. The identifier everyone else asks for.
Most LinkedIn pages the assistants cite belong to people, not companies. The Company Page still matters because every other platform asks for it and Wikidata stores its id. Fill in the name, the one-line tagline, the longer description, the website, the industry, the size, the headquarters and the founding year, and use your category words as the specialties. In Reach, connect the page so it can publish there.
4. G2. The page the assistants quote.
For questions that start with best or top, G2 is cited more often than any vendor’s own page, and what gets quoted is the reviews. Get listed from the vendor menu with a work address on your domain. Pick two categories at most, paste the same descriptions, the logo and the pricing line, and claim the profile when G2 mails you. The free tier is enough.
Then ask ten customers, one at a time and by personal mail, for a review each. G2 checks reviewers against LinkedIn and removes reviews from your own staff. When the listing exists, paste its address into the presence card; the card checks that it still answers.
5. Capterra. One listing, three sites.
Capterra, GetApp and Software Advice share one vendor listing and a more European readership. Sign up with a work email, choose your categories, reuse the same texts and logo, and add five screenshots. Capterra runs free review-collection campaigns; join one instead of asking the same customers twice.
6. Product Hunt. Only with a real launch.
Assistants cite Product Hunt for questions about new tools. A launch with nothing new to say ranks nowhere and the page stays up forever, so wait for a release worth pointing at. Then: a maker account, the same tagline and description, five screenshots, and a first comment in your own words about why the thing exists. Tuesday to Thursday, shortly after midnight Pacific.
7. Your own site. Schema, sameAs, and named authors.
The links from your site to the listings are what lets an assistant confirm that six listings are one company. Your home page needs an Organization block in JSON-LD with the name, the website, the logo, and a sameAs list that carries the LinkedIn page, the Wikidata item, the G2 page and the Capterra page. Most SEO plugins print the block; the sameAs list is the part people forget. Add the founding date and the founder only when the site states them.
Give every article a named author with a role and a bio page, with Person markup on the bio and an author link on each article. Assistants weight named authorship. If your names are not public yet, that is a choice, and the presence card will say so until it changes.
8. Roundups. The best of and alternatives pages.
For comparison questions the assistants cite roundup pages, and a place on one is worth more than any page of your own. The Citations view lists the pages that already have the assistants’ attention. Take the three most cited for your category, find the author, and write a short personal mail: who you are in one sentence, what their list lacks that you do, your G2 page, and one customer story if you have one. No template. Each mail names the page and something specific in it.
Order and time.
Wikidata (twenty minutes), then the sameAs list on your site (ten), then the Google Business Profile (half an hour plus a verification wait), then G2 and Capterra (an hour each), then the roundup mails as an ongoing habit, then Product Hunt when there is a launch. About four hours of one person’s time, spread over a week because of the waits. The presence card in AI Visibility checks all of it again every week without spending a task.
Questions worth asking
Why does the card say Wikidata is missing when an item with my name exists?
Because the check matches on the website, not the name. An item that carries your name but points at another site is a different company, and the card names it so you can write a description that separates you from it.
Does any of this cost tasks?
No. The presence check is a read of public records and your own site, run for you every week and on demand a few times an hour. Nothing here calls a language model.
Should I write a Wikipedia article?
No. Wikipedia requires independent press coverage and does not allow writing about your own company. An article created now is deleted within days and the deletion is public. Wikidata has no such rule and is the record the assistants read.
Last reviewed: September 2026